Episode 5 of The Edge of Now started with one of those headlines that usually makes me bristle: eight executives out in an AI-driven shakeup.

The familiar story writes itself. AI arrives. Heads roll. Another company cuts its way toward an automated future. People become the unfortunate aftermath of a technology nobody can control.

Except that was not the story I found when I read the internal memo. The more interesting story was an organization changing how decisions get made, how people orient around products, where judgment and accountability sit, and how quickly teams can move. The company was not simply removing roles because a machine could perform them. It appeared to be reorganizing around a different way of working.

AI should come for the org chart. It should force every organization to ask whether its structures, workflows, leadership layers, and measures still make sense. What it should not do is become a convenient excuse for cuts an organization has not earned.

Stop Writing the Story as Something Happening to Us

I am tired of job-loss headlines that frame AI as a force acting upon helpless organizations. That framing removes the most important part of the story: human agency.

In this case, the company appeared to study the work and design its way toward a new structure. It created smaller squads, clarified ownership, reconsidered decision-making, and reorganized around the outcomes it needed to produce. That is not an accidental aftermath. It is an intentional operating decision.

Leaders should be willing to say what they are changing, why the business requires it, what evidence informed the move, and how people will be affected. “AI made us do it” is not a strategy. It is an abdication of responsibility.

We still hold the pen. We are not merely co-authors of what technology does to work, organizations, or society. Technology is clay, and we are responsible for what we shape with it.

The Business Problem Has Moved

Expedia is also an important signal because travel is exposed to AI-driven disruption. An intelligent agent can increasingly search, compare, plan, book, and administer a trip. An online booking company cannot assume that the interface and business model that worked in the internet era will remain valuable in an agentic one.

That means the business must reinvent itself. Different capabilities may become more important. Different roles may appear. Engineers may become more central. Decision rights, product ownership, and the relationship with the customer may all need to change.

This is why focusing only on the eight roles that disappeared misses the larger workforce story. The overall organization may grow while the composition of talent changes. The real question is whether the company is building the capabilities required for the next version of its business.

Change must become the strategy rather than something the organization treats as the enemy.

We Are Playing Small and Expecting a Big Story

Another headline showed organizations expressing stronger conviction in AI than the immediate financial returns they can attribute to it. Many people report that AI makes them personally faster, but enterprise performance has not moved at the same rate.

That should force a better conversation about what we are measuring. I hope the average person’s job is not to send email as efficiently as possible or build a PowerPoint deck in 10 percent of the time. Those gains may feel useful, but they do not automatically create business value. Productivity can easily become the art of appearing busy when it should mean producing outcomes the business actually values.

If all we do is make old processes faster, we are playing small. We should be examining what is now possible, how customer experience can improve, which decisions can become better, what new products can exist, and where human capability can expand.

The most powerful technology of our lifetime deserves more imagination than faster email.

What Could Happen Is Not the Same as What Should Happen

The Bill Gates letter Jason brought into the conversation warned that the AI era will be turbulent and that the choices made now will determine whether the good outweighs the bad.

I agree with the central point, but I do not want organizations to interpret caution as a reason to stop. We can govern ourselves into obsolescence if every unanswered question becomes a reason for inertia.

AI literacy should help people move with awareness. It should teach them enough to notice what is happening, interrupt when necessary, ask a better question, understand the tradeoffs, and continue moving. The goal is not a perfect world with no risk. The goal is to know which risks the organization is willing to accept, which boundaries it will protect, and where a human must remain responsible.

A machine may be capable of delivering difficult information, making a decision, or completing an exchange. Capability does not answer whether it should. Care, judgment, taste, trust, and accountability still require intentional design.

Do Not Make Humans the Leftovers

Jason offered a line I have kept thinking about: if we reserve only the work machines cannot do, humans become the leftovers.

That framing reveals how unimaginative the human-versus-machine conversation has become. We keep sorting existing tasks into two piles instead of asking what becomes possible when we redesign the work completely.

Why give the most powerful technology we have seen the same administrative work people already know how to perform? What if we aimed it at problems we have not solved, discoveries we have not made, diseases we have not cured, and capabilities we have not been able to provide?

The original value proposition should never have been that AI can replace people and contract the workforce. It should have been that AI can expand what people and organizations are capable of doing.

Interrogate It Before You Automate It

Every back-to-school season, I tell my children they cannot shop until they purge their closets. Before adding something new, they have to examine what they already own, decide what still fits, and remove what no longer deserves the space.

Businesses should apply the same discipline. Before automating a process, interrogate it. Should it exist? Does it still produce an outcome the business needs? Is the metric meaningful? Does the workflow reflect how value should move now, or are we preserving it because that is how work moved through the old system?

This should be a massive shedding process. We should remove tasks, workflows, processes, measures, and layers that no longer earn their place. Lifting and shifting old work into a faster system is not redesign. It is adding AI glitter to organizational clutter.

Leaders should earn their way to the headline. Study the work. Define the value. Understand the tradeoffs. Redesign the organization with intention. Then tell the real story about what changed and why.

That is the conversation Jason Averbook and I have in Episode 5 of The Edge of Now.

Listen to the Full Episode

Apple Podcasts: https://tinyurl.com/mr3chtyw

Spotify: https://tinyurl.com/2w359w35

YouTube:

All episodes: https://theedgeofnowwithjasonandjess.buzzsprout.com/2633596/episodes

About Jess Von Bank

Jess Von Bank brings more than 20 years of experience across recruiting, talent strategy, employer branding, HR technology, and workforce transformation. She began as a recruiting practitioner before moving into leadership roles focused on bringing workforce solutions to market, including global HR transformation and technology advisory work at Mercer. She is now co-founder of Now to Next, where she combines industry expertise, community building, storytelling, and a human-centered perspective on change.

Leave a comment

Episode 4 of The Edge of Now begins at a familiar turning point. Jason is teaching his first classes of the semester. I am about to have all three of my daughters in high school for the first and only time. Schools are welcoming students back while debating whether phones belong in the learning environment at all.

The policies make sense on one level. Teachers are fighting for attention. They want eyes up, headphones out, and screens down when another human is speaking. They do not want a device to become a distraction, a crutch, or a shortcut around the work of learning.

But removing the technology entirely may solve the wrong problem. Students will need to know how to use these tools with judgment and purpose. The real design question is not simply whether a phone is allowed. It is when the device supports learning, when it interrupts presence, and whether anyone has clearly explained the difference.

That same question now belongs inside every organization.

A Policy Is Not the Same as an Intention

We often reach for rules when what we have failed to create is a shared intention. A ban is easier to communicate than a thoughtful agreement about when a tool is useful, what behavior the environment requires, and what everyone is trying to accomplish together.

I understand why a teacher wants full attention during a lecture. I also understand why a device can become useful when the activity shifts to research, exploration, or creation. The point is not to let everyone do anything at any time. The point is to design the environment around its purpose.

Workplaces need the same clarity. A blanket return-to-office policy does not manufacture connection, just as a remote work policy does not automatically destroy it. Putting people in the same room can still produce a table full of open laptops and divided attention. Connecting everyone through software can still leave people isolated, misunderstood, or reluctant to have a real conversation.

Intentionality asks more of us than location or access. It asks what kind of interaction this moment requires and what behavior will help create it.

Platforms Have Their Own Incentives

We cannot expect platforms to make every decision in the best interest of our humanity. Social media platforms were designed to hold attention because attention supports their commercial model. Notifications pull us back. Features invite us to share more. The system learns what keeps us engaged and gives us reasons to return.

Chatbots introduce an even more intimate version of that dynamic. People increasingly use AI not only to complete tasks but also for counseling, emotional support, and conversations they hesitate to have with another person. The system can feel responsive, available, patient, and personal. That makes it useful, but it also makes understanding the design and incentives behind it more important.

AI literacy cannot stop at learning how to prompt or judge an output. People should understand why platforms seek attention, trust, images, personal information, and continued use. They should know enough to make conscious tradeoffs about what they share and which relationships they allow the technology to simulate.

The platform may be capable of creating a sense of intimacy. That does not mean it can decide what a healthy human relationship should be.

The Human Intervention Is Ours

When Jason described an AI workflow, he pointed out that the system never pauses to say, “Now would be a good time to use your judgment,” or, “Talk to a coworker and get their perspective.”

I am not sure I want the platform making that decision for us. I want people and organizations to retain enough agency to know where a human belongs in the loop and why. In fact, flip it. AI in the human loop, not the other way around.

This requires systems thinking. It also requires us to begin with the experience and outcome rather than the features available inside the technology. How do we want people to feel? What are we trying to accomplish? Which measures still matter? Where is context essential? When should a machine complete the exchange, and when would automation remove the exact conversation through which trust or judgment develops?

The future of implementation is not merely turning knobs and pulling levers until the technology performs. It is listening carefully enough to understand the real problem, questioning whether the old metric still tells the right story, and designing human and machine systems together.

Connection Is Built Through Small Choices

Intentional human design does not always require a grand framework. Sometimes it looks like a meeting leader refusing to begin until every laptop lid is down. If people belong in the meeting, they should be present. If they can continue doing unrelated work without missing anything, perhaps they did not need to be there.

Sometimes it looks like a short team scrum where people say hello, share what is happening, and understand how the day or week is taking shape. Not everyone has to attend every time, and the work does not stop when one person is unavailable. The ritual simply creates a reliable moment of contact and shared awareness.

Sometimes it is deciding whether a message should be a text or a conversation. The easier channel is not always the right one. Screens can give us an excuse to say things we would never say face to face, avoid nuance, or send a difficult thought and consider the interaction complete.

These choices may seem small, but culture is made from repeated behaviors. Connection is, too.

Presence Is a Capability

I have become more intentional about keeping my phone down when I am moving through the world. If I am traveling somewhere new, I want to look at the city rather than the screen. I want to talk with the driver and see what happens in a conversation that was not optimized or scheduled.

I asked my children to order for themselves in restaurants from the time they were young. They had to look at the server, acknowledge another person, answer a question, and articulate what they wanted. Those are ordinary behaviors, but they are also human capabilities. They teach us to engage with someone who is physically present instead of retreating into the device in our hand.

This is not only a concern about young people. All of us need to remember how to engage. Technology makes it possible to move through more of life without doing so, which means presence must become a conscious practice rather than something the environment automatically provides.

We Are Still Holding the Pen

Jason and I do not believe the answer is to reject technology. Bots can talk to bots when that exchange creates value and does not require human judgment. Devices can support research, learning, communication, creativity, and access. AI can make work and life more capable, fulfilling, and beautiful.

But none of those outcomes arrives automatically, certainly not as a guaranteed result of a product roadmap. Humans remain responsible for deciding when people should be involved, what must stay human, and which forms of connection are worth protecting. We still have to design the policies, rituals, workflows, experiences, and boundaries that shape how technology enters our lives.

Being connected is easy. Building connection requires intention.

That is the work in front of us, and it is the conversation at the center of Episode 4 of The Edge of Now.

Listen to the Full Episode

Apple Podcasts: https://tinyurl.com/yujtzh68

Spotify: https://tinyurl.com/ye4hnm8d

YouTube:

Follow The Edge of Now and find every episode here:

https://theedgeofnowwithjasonandjess.buzzsprout.com/2633596/episodes

About Jess Von Bank

Jess Von Bank is a globally recognized voice in HR, talent technology, AI, workforce transformation, and the human experience of change.

She is the co-founder of Now to Next and brings together practitioner insight, storytelling, systems thinking, and a deeply human view of how technology reshapes organizations and people.

Leave a comment

Organizations have spent years connecting people through platforms, devices, chat, video, and increasingly intelligent systems. Yet being connected is not the same as building connection, and technology will not decide for us when people need presence, conversation, judgment, or trust.

In Episode 4 of The Edge of Now, Jason Averbook and Jess Von Bank explore what it means to design humanity into technology-enabled work. Beginning with back-to-school phone policies and moving through AI workflows, remote work, meeting rituals, software design, social platforms, and AI companionship, they examine where human interaction is being weakened, where technology can strengthen it, and who is responsible for drawing the line.

The conversation raises an urgent leadership question. As AI becomes more capable and platforms become more persuasive, will organizations intentionally design the moments when humans should engage with one another, or will connection remain an accidental byproduct of systems built for efficiency and attention? The opportunity is not to reject technology. It is to combine human and machine systems in ways that make work, learning, and life more capable, fulfilling, and human.

Highlights

• Why being connected through technology is not the same as building human connection

• What back-to-school device policies reveal about our struggle to combine attention and technological capability

• Why banning tools outright can deny people capabilities they will need in the future

• How leaders can design specific moments for presence, judgment, conversation, and collaboration

• Why platforms cannot be expected to protect every human interest when their commercial incentives point elsewhere

• What systems thinking and design thinking contribute to AI-enabled work

• Why forward-deployed engineers need to understand human outcomes, not merely technical requirements

• How measures such as quality of hire can become meaningless when the business problem has moved

• Why the next generation of system implementation must integrate human and machine systems

• How technology can become an excuse for behaviors people would not choose face to face

• What meeting rituals, team scrums, and communication choices teach us about intentional connection

• When bots should talk to bots and when humans must remain involved

• Why some members of Gen Z are turning to AI for emotional support and personal issues

• What leaders, educators, parents, and platform users need to understand about persuasive design, privacy, and trust

• Why humans remain responsible for designing technology that makes work and life more fulfilling

Stay Connected:

Jason Averbook
LinkedIn: https://www.linkedin.com/in/jasonaverbook/
X: https://x.com/jasonaverbook
Substack: https://jasonaverbook.substack.com/

Jess Von Bank
LinkedIn: https://www.linkedin.com/in/jessvonbank/
X: https://x.com/jessvonbank
Substack: https://jessvonbank.substack.com/

Now to Next

Website: https://nowtonext.ai

LinkedIn: https://www.linkedin.com/company/now-to-next-ai/

Instagram: https://www.instagram.com/nowtonext.ai/

Now to Next - The Feed: https://www.youtube.com/@nowtonextai

Music licensed through Soundstripe.
Code: MDR2MFWYVNVIS7KB, PK2PQ2D65ENBEQDL

Episode 3 of The Edge of Now started with a familiar idea: back to school. But Jason and I quickly realized that this conversation is no longer just about students, teachers, or classrooms. We are all going back to school right now because AI is forcing every one of us to reconsider what we need to know, how we learn, and which human capabilities become more valuable when information and answers are available almost instantly.

My oldest child is beginning her senior year of high school, which means the inevitable question is getting louder: what are you going to do?

It sounds reasonable because we have spent generations treating education as a path toward a knowable destination. Learn these subjects, earn this degree, choose this profession, and prepare for a future that should unfold in something close to a straight line.

That future was never as predictable as we pretended, and AI has made the uncertainty impossible to ignore. The tools are changing, tasks are changing, and knowledge that once required years of access can now appear in seconds. Asking a young person to know exactly what they will do feels less useful than helping them understand who they are, what they value, and how they will think when the conditions change.

That is not a softer educational goal. It may be the most practical preparation we can give anyone right now.

We Are All Going Back to School

This back-to-school season is not only for students. Leaders, employees, parents, and educators are learning at the same time because there is no complete guide for how people should use these systems, what human work becomes more valuable, or how we should protect agency while taking advantage of what the technology can do.

We are going to be clumsy as we work through it, and that is acceptable. What is not acceptable is pretending the rules are clear when they are not, or asking people to wait for perfect guidance while the world changes around them.

I do not love the language of inevitability. Saying that AI is simply happening to us can become an excuse to surrender the pen and accept every consequence as unavoidable. I believe we are still holding the pen on some very important decisions, including how systems are designed, what human value they serve, how people participate, and who benefits.

The goal should be to make change happen for people and with people, not merely to them. That requires enough understanding, experimentation, and self-education to participate in shaping the environment rather than waiting for a manual that may never arrive.

Humans Have to Win Here

I believe systems can and should be designed for ultimate human value. People need to come out ahead in what we build, or we need to ask what we are building for.

That does not mean every transition will be painless or that nothing should change. It means human value cannot be treated as a hopeful side effect of technological progress. It has to become a design principle.

If information and basic production are abundant, the capabilities that become more valuable are profoundly human. Expertise matters, especially when it is combined with curiosity, humility, hunger, lived experience, and contextual judgment. Someone who has spent years learning what good work looks like does not become obsolete because a system can generate an output. Their ability to question, apply, interpret, and recognize quality becomes more important.

We do ourselves a disservice when we frame this as young AI-native people on one side and experienced workers on the other. Pair the person who moves naturally with the technology with the person who knows what good looks like. Let them learn from each other. The multiplication happens in the combination.

Use the Tool, but Do Not Look Like You Used It

We are creating a deeply confusing environment for learning. Platforms first told everyone they were behind and offered AI assistance inside the product. Now some of those same environments are introducing detection systems, reporting features, labels, and watermarks.

The message becomes impossible to follow: use AI, but make sure your work does not look like you used AI.

That contradiction is not a minor communications problem. It leaves students, employees, writers, and creators without a social contract. They do not know which forms of assistance are acceptable, what transparency requires, how authorship should be understood, or when using a learning tool becomes cheating.

Detection alone does not solve any of those questions. We need to ask what we are actually policing. Is there a human idea? Was judgment applied? Are the sources credible and attributed? Can we see evidence of thinking, accountability, and care?

A document should not become worthless because someone used a system to proofread or improve it. At the same time, a stream of unexamined machine output should not be mistaken for human thought simply because it passes a detector. Quality, judgment, authorship, and responsibility deserve a more serious conversation than a label can provide.

Learning Should Make Us Better at Hard Things

I think about endurance training when I think about durable human capability. You do not train because hard things become easy. You train because you become better at doing hard things. You learn how to manage your resources, respond to discomfort, and continue when the conditions change.

Learning should work the same way. It should not merely deliver an answer. It should strengthen the ability to ask a better question, follow a line of reasoning, recognize a weak argument, and arrive at an informed conclusion.

One experimental learning model Jason and I discussed did not provide users with answers. It responded with questions, and people hated the experience. Their cognitive ability and the quality of their outputs improved, even though the process felt less convenient.

That tension matters. If every system is designed to remove effort, we may also remove the productive struggle through which judgment develops. A useful learning tool should assist us without quietly replacing the thinking we need to practice.

Presence Is a Capability Too

Curiosity and critical thinking require something we rarely protect: presence. People need time to sit with a question, listen to one another, notice what is happening, and allow a thought to form before another notification interrupts it.

Some of the most valuable learning experiences are carefully created containers rather than content modules. A teacher arranges the room so people can see one another, removes distractions, asks a meaningful question, and leaves enough open space for someone to finally say what they are thinking.

Schools and workplaces should create more of those spaces. Technology can be used intentionally as a research or creative tool, but constant access is not the same as learning. Sometimes the most valuable design choice is to remove the device and restore attention to the people in the room.

Identity Is the Anchor

We keep asking young people what they need to know and what they plan to do. Adults are asking themselves the same questions about careers, skills, and relevance. Those questions matter, but they are downstream from something more durable.

You need to know who you are and what you value. That becomes the standard against which new tools, social pressure, institutional rules, and easy answers can be tested. It helps you remain open without swaying every time the market changes direction.

AI can offer options, generate ideas, summarize information, and help us learn. It cannot take responsibility for the person we become or the values we bring to the choices in front of us.

The answer is not the destination anymore. The deeper work is becoming the kind of person who knows what to do with it.

Listen to the Full Episode

Apple Podcasts: https://tinyurl.com/3btuew4j

Spotify: https://tinyurl.com/484kdt4r

YouTube:

Follow The Edge of Now and find every episode here:

https://theedgeofnowwithjasonandjess.buzzsprout.com/2633596/episodes

About Jess Von Bank

Jess Von Bank is a globally recognized voice in HR, talent technology, AI, workforce transformation, and the human experience of change.

She is the co-founder of Now to Next and brings together practitioner insight, storytelling, systems thinking, and a deeply human view of how technology reshapes organizations and people.

Leave a comment

Share Jess Von Bank | Co-Founder, Now to Next

AI has made information easier to access and first drafts easier to produce, but abundant answers do not automatically create better thinkers. As schools, universities, and workplaces race to introduce AI, the more urgent question is what people should learn when the answer itself is no longer scarce.

In Episode 3 of The Edge of Now, Jason Averbook and Jess Von Bank explore why everyone is going back to school, regardless of age or role. They examine the durable human capabilities that become more valuable as technology accelerates, including curiosity, judgment, adaptability, presence, expertise, and the ability to ask better questions.

The conversation moves beyond prompt engineering and AI adoption to a more consequential challenge: how do we create experiences that strengthen thinking, preserve human agency, and help people practice their way into new capabilities? For leaders, educators, parents, and anyone trying to navigate this moment, the goal is not simply to know more. It is to become a better learner in a world that will not stop changing.

Highlights:

• Why back-to-school season now applies to leaders, employees, parents, students, and educators

• Why waiting for a dramatic AI threshold distracts us from changes already happening around us

• How organizations can help change happen with people instead of merely to them

• Why human value must be designed into the technological revolution from the beginning

• Which durable capabilities become more valuable when information and basic production become abundant

• Why expertise matters most when it is paired with curiosity, humility, adaptability, and contextual judgment

• How practice, feedback, reflection, and real work build capabilities that a conventional course cannot

• Why the future of learning may be less about transferring knowledge and more about creating better thinkers

• How AI detection tools and unclear rules can punish people for using the very tools they were told to adopt

• Why prompt engineering should not become the center of organizational AI education

• What organizations should measure instead of logins, usage, and superficial adoption

• Why leaders must define value and outcomes before prescribing tools

• How schools and workplaces can create the capacity, culture, and presence required for meaningful learning

• Why knowing who you are may matter more than knowing exactly what you will do

Stay Connected:

Jason Averbook
LinkedIn: https://www.linkedin.com/in/jasonaverbook/
X: https://x.com/jasonaverbook
Substack: https://jasonaverbook.substack.com/

Jess Von Bank
LinkedIn: https://www.linkedin.com/in/jessvonbank/
X: https://x.com/jessvonbank
Substack: https://jessvonbank.substack.com/

Now to Next

Website: https://nowtonext.ai

LinkedIn: https://www.linkedin.com/company/now-to-next-ai/

Instagram: https://www.instagram.com/nowtonext.ai/

Now to Next - The Feed: https://www.youtube.com/@nowtonextai

Music licensed through Soundstripe.
Code: MDR2MFWYVNVIS7KB, PK2PQ2D65ENBEQDL

We even lured everyone in with snake-charming AI features:

“Want help writing?”
“Enhance your post.”
“Write with AI.”
“Make it shorter.”
“Make it funny.”
“Write a hook.”
“Make it casual.”
“Brainstorm titles.”
“Create a summary.”

We embedded these prompts inside every text editor on every platform, the technological equivalent of someone standing over your shoulder whispering, You know, the machine could do that for you. How devilish.

Just as quickly, we rolled out detection features with all our authenticity policing (I spy an em dash, you cheater), status signaling (I’m a real writer, not one of those fake types), aesthetic disgust (ewww, this is so formulaic, writing should be messier).

So, a couple of observations.

We’ve stuffed so much AI down everyone’s throats, we’re all coming to detest it. Myself included. Too much of a good thing is still too much, no matter how good. Or, what do they say, the dose makes the poison?

But the hypocrisy of this moment is damaging, disrespectful, and defeating.

The first era of AI was organized around the fear of falling behind.

The second era, the one we’re squarely in now, is organized around the fear of being found out.

I wish it were as simple as hypocrisy. It runs deeper.

This is what happens when several unresolved social conflicts collide:

• Institutions coerced adoption without creating a social contract for acceptable use.

• Technology companies treated the world’s creative work as available training material, then transferred responsibility for ethical use to individual users.

• Employers celebrated AI productivity while threatening the economic value of the people expected to produce it.

• Platforms filled themselves with cheap synthetic content, then invited users to police one another for the degradation those platforms economically incentivized.

• People were told AI literacy would confer status. Now visible dependence on AI can cost status.

• We never established where assistance ends and authorship begins, so everyone is making up a moral boundary and prosecuting everyone else from it.

AI spotting is becoming a status ritual

That last conflict, where assistance ends and authorship begins, matters quite a bit.

AI-shaming isn’t simply an authorship dispute. It’s becoming a class and competence ritual. “I can spot AI” has become a way to claim taste, discernment, and judgment (all our favorite new words, have you noticed?) over someone presumed to lack the qualities of the thinking class.

There’s a name for part of the reflex underneath it: the effort heuristic. We tend to judge the value of work by the effort we believe went into producing it. Sweat is data. And AI definitely breaks that calculator.

When a decent paragraph can appear from a single prompt, something in us panics. Mockery becomes the fastest way to restore the old scoreboard:

I toiled, therefore my work is worth more than yours.

It isn’t really about the sentence. It’s about protecting the price of effort itself.

But effort has never been the same thing as value. A person can spend six hours writing something terrible. Another can produce something extraordinary in twenty minutes. We’ve all sat through the meeting someone worked very hard to prepare.

Work doesn’t become meaningful merely because it hurt (insert rock ballad). And AI-assisted work does not become meaningless merely because it didn’t.

The question should be whether a human being brought an idea, exercised judgment, verified the claims, made deliberate choices, and remained accountable for what was published.

Instead, we’re inspecting the prose for machine fingerprints and missing the almighty point of the thing.

Human error is becoming a CAPTCHA

Here comes an irony so ironic I just can’t even:

Now, NOW, we’re beginning to deform our natural writing to perform humanness.

Has anyone else removed an em dash you’ve literally always used? Made sentences deliberately less orderly? Smashed paragraphs together to avoid a telltale stack of single sentences? (I’m doing it right now.) Left imperfections behind as proof of life Even, EVEN, asked AI to “make this not sound like AI”?

Human error is becoming a CAPTCHA.

We are deliberately writing worse, or at least less naturally, to prove we wrote anything at all.

For Chrissake, what are we doing here?

I wish aliens would come down to Earth (who are we kidding, they’re probably already here) and tell us how utterly inane we are.

Subscribe now

What are we actually objecting to?

What is the point of this morality exercise, anyway? What, exactly, are we objecting to? Low-quality content? Content published without genuine thought? Ideas taken without attribution? Undisclosed AI assistance? Confidently stated falsehoods? Manufactured content to please the algo gods?

These are not the same ethical failure. And none of these problems are unique to AI.

The internet was full of human slop well before AI entered the picture. Humans have always published derivative thoughts, stolen ideas, produced formulaic writing, pretended to know things they didn’t, and manufactured empty content to appease an audience or algorithm.

AI made it cheaper, faster, more abundant, more polished. And remember, we told people this was an amazingly good thing.

Sure, it might be consequential. But abundance is not the same as authorship, polish is not ill-willed deception, and assistance is not automatically fraud.

When we call all of it “slop,” we avoid making the distinctions that would actually help:

Was there a human idea?
Did the author exercise judgment?
Can they defend the argument?
Did they verify what they published?
Did they attribute what belonged to someone else?
Did they contribute anything, or merely manufacture another object for the feed?
Does disclosure materially matter in this context?

And maybe my stupidest question, posed in all seriousness: Is it our job to decide whether someone else’s idea required enough effort to deserve expression?

We never defined worthy use

What we’re finally sensing is that everyone was told to use AI without any definition of what worthy use looked like.

We were told to become more productive without a definition of value.

We were told to get on board with zero idea of the destination.

We turned adoption into a game of winners and losers. Either you were doing it or you were behind. You were self-teaching your way toward expertise in everything, or you were a Luddite. Employers bragged about eliminating jobs and slowing hiring because AI could do the work, while workers were expected to figure it out, save themselves, sink or swim.

Everyone had to evolve faster than the person next to them so they wouldn’t be chairless when the music stopped.

Then, once they used the tools built, marketed, and embedded for precisely that purpose, we began shaming them for it.

Damned if they do. Damned if they don’t.

That’s why I said we were building the Titanic in plain sight. We gave people a tool, a threat, and an adoption target—but no covenant.

No shared understanding of authorship.

No definition of valuable use.

No agreement about disclosure.

No distribution of the gains.

No protection from the risks.

No coherent story about the human future we were supposedly so excited to build.

And now we’re asking individual users to settle all of that by reporting one another.

We’ll probably monetize humanness next

Here’s my cheeky prediction: give it eighteen months.

The same platforms currently congratulating themselves for cracking down on AI slop will start selling us the cure.

“Verified Human” badges.

“100% Human-Made” premium tiers.

A special checkmark certifying that your thoughts are artisanal, organic, locally sourced, and untouched by large language models. “Pure” (cringe).

A predictable platform playbook is to create or amplify a behavior, scale it until users revolt, and then turn the remedy into a product. First, sell us AI assistance. Then, sell us protection from AI abundance. Finally, sell us proof of our own humanity.

Don’t be shocked when “human-made” appears as a checkout option. ;) Someone remember I said this so I can say I toldja so.

For the love of all that is holy, stop the whiplashing.

This is not merely a content quality problem. It is a spectacular social design failure.

Leaders, step up.

If your AI enablement, adoption, or transformation program cannot explain what worthy use looks like—where assistance ends, where accountability begins, how value is defined, and how people benefit—you do not have an AI strategy.

You have tools, pressure, and ambiguity.

We can do better in this moment.

In fact, our next webinar touches quite a lot of this. If your organization is trying to untangle adoption from value—and AI use from actual transformation—we’d like to help.

Not me selling you anything, just me begging us to be less inane.

Register to attend live or get the replay.

Godspeed,

Jess Von Bank

Co-Founder, Now to Next

I write at the intersection of work, technology, and humanity. You’ll find essays here that challenge orthodoxy and make space for possibility. All are written with the future—and the people who will live in it—in mind.

Subscribe now

P.S. Please don’t come for me. I hate actual AI slop as much as anyone. But just, everything I said above. Wild times to be a human, amiright?

.

AI can preserve organizational knowledge, expand human capability, and capture context that traditional systems routinely miss. It can also record, retain, and repurpose more of our work, behavior, and judgment than most people fully understand.

In Episode 2 of The Edge of Now, Jason Averbook and Jess Von Bank examine the complicated relationship between AI, privacy, trust, and human value. They explore what happens when organizations encourage experimentation while limiting access, when useful technology begins to feel like surveillance, and when a person’s knowledge becomes training material for a system.

The question is not whether organizations should use this technology or protect people from its risks. The real challenge is designing governance that allows responsible progress while ensuring that the people whose knowledge creates value are recognized, protected, and included in that value.

Listen to the conversation and consider what a fair value exchange between people, organizations, and AI should look like.

Highlights:

• Why encouraging AI experimentation while restricting tokens sends a contradictory message

• How organizations can support different levels of AI participation without rebuilding an outdated center-of-excellence model

• Why probabilistic technology requires new education and clearer expectations

• When meeting recorders, wearables, and voice technology begin to feel like surveillance

• What separates listening, recording, storing, and repurposing human information

• Why trust requires more than accepting a privacy policy or checking a consent box

• How AI could preserve tacit knowledge without treating employees as extractable training data

• Why effective governance should keep responsible experimentation moving

• How voice can capture context that forms, fields, and surveys routinely miss

• What organizations owe people when their experience and judgment help train intelligent systems

• How AI could make previously invisible human contribution easier to recognize and reward

• Why human agency must be designed into the value exchange from the beginning

Stay Connected:

Jason Averbook
LinkedIn: https://www.linkedin.com/in/jasonaverbook/
X: https://x.com/jasonaverbook
Substack: https://jasonaverbook.substack.com/

Jess Von Bank
LinkedIn: https://www.linkedin.com/in/jessvonbank/
X: https://x.com/jessvonbank
Substack: https://jessvonbank.substack.com/

Now to Next
Website: https://nowtonext.ai/

Music licensed through Soundstripe.
Code: MDR2MFWYVNVIS7KB, PK2PQ2D65ENBEQDL

OpenAI just published research showing that 43.5 percent of occupation-specific ChatGPT use crosses traditional job boundaries. For HR, it was 69 percent.

The report is called How AI Is Expanding What People Do at Work.

Of course it is.

Every research frame has a point of view. This one also has a commercially viable business model.

If workers have always learned sideways, crossed functional boundaries, borrowed expertise, and routed around bad work design, ChatGPT is a useful new tool appearing inside a very old human behavior.

If ChatGPT is causing workers to stretch beyond their jobs, busting up occupational boundaries and reorganizing the division of labor, then AI becomes the transformational force every enterprise urgently needs to purchase, integrate, govern, and scale.

One interpretation credits people.

The other sells enterprise AI.

So, behold: after decades trapped inside the boxes of work, the model has arrived to stretch people east-west, liberate them from their job descriptions, and unlock capabilities they apparently never possessed before.

You’re welcome. Signed, OpenAI.

Except the research does not prove that.

Subscribe now

What the research actually found

OpenAI analyzed more than 800,000 work-related messages from U.S. ChatGPT users. It found that 16.8 percent of all work-related messages—and 43.5 percent after generic work such as writing, summarizing, and scheduling was removed—concerned tasks historically associated with an occupation other than the user’s own.

OpenAI calls this “task crossover.”

But the study did not establish that these workers had never performed those tasks before. It did not observe whether the work was already part of their real job, whether someone else would have handled it without AI, whether the employee was preparing for a specialist handoff, or whether the output was ever used.

The report itself says its findings are descriptive. It cannot tell us how the same people would have allocated their work without AI.

So “AI is expanding what people do” is not a finding.

It’s an interpretation, one that awards a remarkable amount of agency to the product being sold by the company conducting the research.

Commercial self-interest does not make the data false. It makes the story wrapped around the data worth interrogating.

Follow the commercial interests. Then draw your own conclusions.

As you might imagine since I dedicated an entire essay to the research, I have a few thoughts of my own:

1. People have always subverted work design

AI is not expanding what people do.

I’d like to give people more credit, please.

Any Chief Learning Officer could tell you what I’m about to tell you: people have always subverted work design.

Not necessarily as an act of rebellion. Usually as an act of necessity.

Work as designed is never work as encountered.

The process map cannot anticipate every exception. The job description cannot contain every capability. The org chart cannot account for every relationship through which work actually moves. The formal curriculum cannot possibly teach everything someone will need to know in the moment they need to know it.

So people learn sideways.

They borrow expertise. They watch someone else. They build workarounds. They call the person who knows the person. They open a spreadsheet no one approved, find a tutorial, ask a colleague, join an unofficial group chat, and teach themselves enough to solve the problem in front of them.

They cross functions because the work crosses functions.

They operate outside their job descriptions because job descriptions are incomplete representations of work, written at a particular moment, for administrative purposes, by people who cannot anticipate every problem the job will encounter.

People did not stay obediently inside occupational boundaries until ChatGPT arrived and set them free.

Any decent learning leader knows the organization does not formally teach most of what the organization knows. People do. Capability travels through peers, stretch assignments, observation, experimentation, proximity, necessity, and the moment someone says, “I’ve never done this before, but let me see if I can figure it out.”

We have names for this: informal learning, social learning, experiential development, job crafting, communities of practice.

When it produces the outcome the organization wanted, we call it agility and innovation.

When it makes the organization nervous, we call it shadow behavior and noncompliance.

One thing it’s not: new.

2. AI did not create the cow paths

Many roads began as paths worn into the ground through repeated use. Animals walked them. People followed. Wagons widened them. Eventually, someone paved them.

The road didn’t create the route.

The route showed us where the road belonged.

Organizations routinely forget this. We draw the map, define the roles, build the processes, assign the permissions, and then act surprised when people cut across the grass to get the work done.

Those are the cow paths of the organization: unofficial networks, missing handoffs, borrowed expertise, improvised tools, and work that travels through people because the formal operating model cannot carry it.

AI did not create those paths.

It gave us an aerial view of them.

OpenAI is looking at the worn grass, comparing it with the official road map, and concluding that the technology caused people to leave the road.

But perhaps the road was never where the work needed to go.

Perhaps the occupational map was wrong.

Perhaps employees have been compensating for flawed work design all along.

AI may be less like the source of a new river and more like tracer dye added to water already flowing. It lets us see the channels, diversions, and underground routes that were difficult to observe before.

We should not watch the dyed water and credit the dye for creating the river.

3. The formal role came from work. The actual work came from somewhere else.

The most interesting detail in the report is its methodology.

OpenAI identified users’ occupations through self-reported department and role information connected to their ChatGPT Business accounts. But the work-related messages it analyzed came from those users’ individual ChatGPT accounts.

These were individual accounts, not necessarily personal uses. OpenAI had filtered the messages specifically for work-related activity.

The Business account supplied the formal identity.

The individual account supplied the evidence of how the person actually worked.

OpenAI treats the distance between the two as evidence that AI is broadening the role.

I see evidence that the formal role was never a reliable description of the whole person or the work they performed.

The company account told OpenAI which box the employee belonged in. The individual account showed how little the work respected the box.

That is not necessarily AI-driven task expansion.

It may be organizational fiction colliding with behavioral evidence.

This shouldn’t surprise us. Enterprise AI began entering the organization through employees when ChatGPT was released, long before it arrived through procurement, approved architecture, or a formal change initiative.

What should surprise us is that the pattern continues several years later.

Enterprise licenses are flowing. Approved use cases are multiplying. Governance frameworks are hardening. Adoption and enablement initiatives are everywhere. Yet work-related AI behavior is still occurring through individual accounts. And as the crow flies, not as the enterprise is mapping the route.

Enterprise access did not replace the consumer pathway. It was layered on top of it.

So while the formal transformation has begun, the informal one never waited.

4. Employees were already the integration layer

Enterprise leaders tend to think about integration as a technical problem: systems connecting to systems, data moving between platforms, agents invoking workflows, security and permissions controlling the exchange.

But employees have always been the integration layer.

They carry context between systems that cannot share it. They translate across functions that use different language. They reconcile processes designed separately. They remember the exception the workflow cannot accommodate. They know who to call, what to ask, and which part of the official process can be safely ignored when reality refuses to cooperate.

AI did not make employees integrators.

Fragmented organizations did.

What AI may do is make some of that integration faster, easier, more scalable, or more visible. It may reduce the cost of attempting an unfamiliar task. It may allow someone to travel further down an adjacent path before requiring a specialist.

That matters, but acceleration is not origination, and assistance is not authorship.

The person still recognizes the need, supplies the context, connects the disciplines, evaluates the response, exercises judgment, and decides what happens next.

The model is not traversing the organization.

The employee is.

5. HR is not borrowing everyone else’s job

The 69 percent figure attached to HR is especially easy to sensationalize.

But what exactly did we think HR was?

HR already operates at the intersection of regulation, finance, technology, communications, analytics, organizational design, psychology, operations, and business strategy. It is one of the most cross-functional areas in the enterprise.

A high rate of “task crossover” may not mean AI is turning HR professionals into something new. Or that HR is dead, the headline I was tempted to throw on my initial quick and dirty analysis.

It may mean the occupational definition of HR was fictionally narrow to begin with.

Tasks do not belong to occupations as neatly as occupational databases suggest. People do not experience work as a tidy inventory of activities owned by mutually exclusive functions. A worker trying to solve a problem does not stop at the border and wait for their passport to be stamped.

HR is not dead, but the box we drew around it might be.

And if we interpret this research as proof of AI’s expansive power, we miss a more uncomfortable possibility: organizations have relied on people to exceed their formal roles for decades without consistently recognizing, developing, rewarding, or compensating them for doing so.

That is not AI transformation.

That is the old operating model, now visible in a new dataset.

6. The shadow operating model

Even “shadow transformation” may give AI too much credit.

What we are observing may be the shadow operating model that existed long before frontier models: the real network of relationships, judgment, favors, workarounds, learning, and invisible labor required to make the formal organization function.

AI did not create that operating model. It became another surface on which the operating model could operate, and one capable of leaving a remarkably detailed record. The shadow is not merely the account. The shadow is the work the organization has always depended on but never fully designed, measured, or valued.

That is the inverted transformation story I think OpenAI’s research actually reveals.

The enterprise thinks it is introducing AI to the workforce. In practice, the workforce is carrying AI into the existing human operating system: the unofficial routes through which work has always flowed.

The technology is following the cow paths, not the other way around.

Why Jason and I are betting on voice

This is also why Jason and I keep betting on voice AI as the future of work.

Hypertasking is the obvious use case. Voice allows people to direct, delegate, retrieve, synthesize, and coordinate while their hands, eyes, and attention are engaged elsewhere.

But speed is not the most interesting part.

Voice is interesting because it has a natural cow-path quality.

Most enterprise systems require people to stop working, enter the system, translate what happened into predetermined categories, and document a cleaned-up version of reality. The system does not capture the work. It captures whatever survived the conversion into fields, forms, tickets, and drop-down menus.

Voice can travel with the work.

People think aloud. They narrate exceptions. They explain what made this customer different, why the official process failed, which workaround they used, and who actually helped them solve the problem. They pause, correct themselves, change their minds, and reveal the ambiguity that disappears when experience is converted into structured data.

A form captures the answer. Voice can capture the story behind it.

That is the truth-telling potential of voice AI. Not that spoken words are inherently more truthful, but that conversation is closer to how people naturally make sense of work. It asks them to perform less translation between lived experience and enterprise record.

Voice does not teach people a new system behavior. Rather, it meets an existing human habit. That makes it capable of following the cow paths instead of forcing everyone onto the official road. It can surface the unofficial handoffs, contextual judgment, institutional memory, and practical knowledge the organization depends on but rarely captures.

Done well, voice AI could give leaders a much richer understanding of how work actually happens without asking employees to become full-time documentarians of their own labor.

Done badly, it becomes the ultimate surveillance layer.

That distinction matters enormously. People will only tell the truth if they understand who is listening, what will be retained, how their words will be used, and whether candor will be rewarded or punished. The moment voice becomes an invisible monitoring system, people will edit themselves, and the cow paths will simply move somewhere else.

Voice AI could become the most human work interface we have built.

It could also become the most invasive.

That is a design and governance choice, not a technological inevitability.

Our bet on voice is not simply that talking will be faster than typing. It is that the future of enterprise AI will belong to systems capable of meeting people where work naturally occurs: inside conversation, context, motion, ambiguity, and human habit.

Voice is the path before the pavement.

Follow the commercial interests

Read from a bird’s-eye view, OpenAI’s research is also a commercial map.

OpenAI does not need to convince enterprises that employees want AI. Employees have already demonstrated that demand. OpenAI can now argue that its product is not merely making existing work faster; it is expanding jobs, collapsing handoffs, and reorganizing the division of labor itself.

That is a much larger value proposition.

If AI is the force busting up occupational boundaries, companies need more than a few licenses. They need enterprise-wide access, integrations, governance, security, training, workflow redesign, and a permanent platform relationship.

OpenAI’s framing turns organic employee behavior into evidence for enterprise architecture.

Again, that does not make the research invalid. It makes the interpretation commercially useful.

Following commercial interests isn’t cynicism, at least for me. It’s honest analysis.

OpenAI, Anthropic, Microsoft, Google, and every other model provider chasing the enterprise are not merely competing to supply a better model. They are competing to become the place where work happens, and therefore the place where work can be observed, organized, governed, and monetized.

The enterprise product does not necessarily introduce AI behavior into the organization. It simply attempts to capture behavior that already exists.

The provider that can see the cow paths has an obvious reason to offer to pave them.

Voice raises the stakes even further. A model that sits inside conversation does not simply process more work. It potentially gains access to a more natural, contextual, and truthful layer of human behavior.

That could make the system extraordinarily useful.

It could also make the commercial and governance interests surrounding it extraordinarily powerful.

The closer technology gets to human habit, the more carefully we should ask whose interests it serves.

Subscribe now

Follow the people

The lesson for enterprise leaders is not to marvel at AI for making workers more capable.

It is to recognize how much capability their existing systems have failed to see.

Go look at the grass.

Where are employees already working beyond their formal roles?

Where are they learning from one another because the curriculum does not reach the moment of need?

Which handoffs have always depended on unofficial relationships?

Who carries context across functions because the systems cannot?

Which workers have quietly become the analysts, translators, troubleshooters, marketers, designers, and policy interpreters on their teams?

Where has resourcefulness become recurring responsibility without a corresponding change in authority, workload, title, development, or pay?

Not every cow path should become a road. Some are unsafe. Some need specialist review. Some expose genuine governance or data risk. Some exist only because the official process is underfunded, inaccessible, or broken.

But the answer is not to bulldoze the paths and order everyone back onto roads that do not lead where the work needs to go.

Study the paths.

Understand why people chose them.

Build infrastructure around how work actually happens—not around the fiction that people ever remained inside their boxes.

This is where HR and learning leaders have an enormous opportunity. Not to defend an increasingly brittle taxonomy of jobs and skills, but to make the real movement of capability visible. To distinguish occasional experimentation from durable role expansion. To recognize who is carrying the operating model across its gaps. To ensure people are prepared, protected, recognized, and paid for the work they are actually doing.

And when voice gives us a more natural way to hear those truths, good God, protect the truth teller.

Do not turn candor into surveillance. Do not turn human adaptability into another source of extraction. Do not capture the cow paths merely to claim ownership of them.

Use what people reveal to build work that serves them better.

That would be radical self-disruption.

It would also give people the credit they deserve.

AI isn’t expanding what people do. People have always expanded beyond what work design imagined for them.

The model did not create the cow paths. It arrived just in time enough to find them, and barely early enough to claim the credit.

Jason Averbook and I launched Episode 1 of a new podcast last week, The Edge of Now. Some of the ideas above are expressed in our inaugural episode, When Work Outgrows the Organization. Take a listen and smash that like button, as my kids would say.

Godspeed,

Jess Von Bank

Co-Founder, Now to Next

I write at the intersection of work, technology, and humanity. You’ll find essays here that challenge orthodoxy and make space for possibility. All are written with the future—and the people who will live in it—in mind.

Subscribe now

AI is changing what work is, but most organizations still rely on job descriptions, structures, and leadership practices built for a less fluid world.

In the first episode of The Edge of Now, Jason Averbook and Jess Von Bank examine why people increasingly contribute across functions while organizations continue to keep them in boxes. They discuss the growing value of organizational context, the difference between routine and frontier problems, and what happens when companies invest in AI without redesigning the work around it.

The capabilities already exist. The harder question is whether leaders are prepared to organize around outcomes, expand human potential, and build organizations that learn as quickly as work changes.

Listen to the conversation and consider what your organization may already have outgrown.

Highlights:

• Why “What do you want to do next?” may be more useful than asking what someone wants to be

• What cross-functional AI use reveals about the changing nature of work

• Why broad learning may be becoming more valuable than narrow expertise

• How outdated job descriptions quietly constrain organizational value

• Why organizational context may matter more than model selection

• What companies should document before introducing agentic workflows

• The difference between routine, complex, and frontier problems

• Why advanced AI should be reserved for questions requiring genuine discovery

• What separates AI investment from meaningful work redesign

• Why leadership may be the greatest barrier to becoming AI-native

About the Hosts

Jason Averbook

Jason Averbook has spent more than 30 years shaping HR technology and workforce transformation. His career includes leadership roles at PeopleSoft and Ceridian, co-founding and leading Knowledge Infusion from 2005 to 2012, serving as CEO of The Marcus Buckingham Company, and co-founding and leading Leapgen from 2018 to 2023. After serving as Mercer’s Global HR Transformation Leader, he co-founded Now to Next to help leaders prepare people and organizations for an AI-shaped world

Jess Von Bank
Jess Von Bank brings more than 20 years of experience across recruiting, talent strategy, employer branding, HR technology, and workforce transformation. She began as a recruiting practitioner before moving into leadership roles focused on bringing workforce solutions to market, including global HR transformation and technology advisory work at Mercer. She is now co-founder of Now to Next, where she combines industry expertise, community building, storytelling, and a human-centered perspective on change.

Stay Connected:

Jason Averbook
LinkedIn: https://www.linkedin.com/in/jasonaverbook/
X: https://x.com/jasonaverbook
Substack: https://jasonaverbook.substack.com/

Jess Von Bank
LinkedIn: https://www.linkedin.com/in/jessvonbank/
X: https://x.com/jessvonbank
Substack: https://jessvonbank.substack.com/

Now to Next
Website: https://nowtonext.ai/

Music licensed through Soundstripe.
Code: MDR2MFWYVNVIS7KB, PK2PQ2D65ENBEQDL

The world doesn't need another podcast. It needs a better conversation.

Welcome to The Edge of Now, a weekly conversation with Jason Averbook and Jess Von Bank exploring the space where technology meets humanity and where the future of work is being shaped in real time.

Each week, they'll unpack the biggest signals shaping our world by asking four questions: What happened? What's the story behind it? What's the human impact? And where is the edge?

If you're looking for thoughtful conversations that move beyond the hype and help you navigate what's next, you're in the right place.

New episodes release every Friday at 7:00 AM Central Time. Subscribe wherever you listen to podcasts, and we hope you'll tune in every Friday as we explore the ideas, questions, and possibilities waiting at the edge of now.

Music licensed through Soundstripe.

Code: MDR2MFWYVNVIS7KB, PK2PQ2D65ENBEQDL

About the Hosts

Jason Averbook

Jason Averbook has spent more than 30 years shaping HR technology and workforce transformation. His career includes leadership roles at PeopleSoft and Ceridian, co-founding and leading Knowledge Infusion from 2005 to 2012, serving as CEO of The Marcus Buckingham Company, and co-founding and leading Leapgen from 2018 to 2023. After serving as Mercer’s Global HR Transformation Leader, he co-founded Now to Next to help leaders prepare people and organizations for an AI-shaped world.

Jess Von Bank
Jess Von Bank brings more than 20 years of experience across recruiting, talent strategy, employer branding, HR technology, and workforce transformation. She began as a recruiting practitioner before moving into leadership roles focused on bringing workforce solutions to market, including global HR transformation and technology advisory work at Mercer. She is now co-founder of Now to Next, where she combines industry expertise, community building, storytelling, and a human-centered perspective on change.

Stay Connected:

Jason Averbook
LinkedIn: https://www.linkedin.com/in/jasonaverbook/
X: https://x.com/jasonaverbook
Substack: https://jasonaverbook.substack.com/

Jess Von Bank
LinkedIn: https://www.linkedin.com/in/jessvonbank/
X: https://x.com/jessvonbank
Substack: https://jessvonbank.substack.com/

Now to Next
Website: https://nowtonext.ai/

Music licensed through Soundstripe.
Code: MDR2MFWYVNVIS7KB, PK2PQ2D65ENBEQDL

Seven weeks ago I opened with a meeting you had probably been in, and a metaphor I knew would make some of you uncomfortable. The Titanic in plain sight. I asked you a question I’ve been carrying through every piece since: what would you do differently if you actually believed the people inside these decisions deserved co-authorship? I told you I would call four witnesses, walk you through the framework that holds them together, and ask the question one more time at the end. We have done all of that.

Now I am asking the question.

But before I do, I want to say one thing out loud. The thing that has been running underneath every piece in this series, that I have not yet stated directly, and that is the whole of it.

We are not automating work. We are automating dysfunction.

That is the verdict. That is what the seven weeks have been building toward. Every piece in this series has been a different angle on the same observation. Véliz showed us that we are using probabilistic systems in places where they do not belong, and treating the outputs as facts. The Amodeis showed us what it looks like when stated values actually constrain behavior, and what it looks like when stated values are absent. Brynjolfsson showed us the data on who is being hurt and how, and named the choice no one was naming. The view from inside the room showed us the Five Wounds of Change, alive in the language of the people carrying them, and the unfinished sentence the executive sponsor holding the pen could not complete. The framework showed us the order the work actually requires — Mindset before Heartset before Skillset before Toolset, the 4SET sequence, in the order almost every enterprise gets backwards. And underneath all of it, in every room this series has taken you into: Hands, Heads, Hearts. What we automate. What we augment. What we amplify. Three questions. Every rollout answers them whether anyone asks or not.

All of it points to the same place. We are pouring the most powerful technology of our generation into operating models that were already broken, decision-making structures that were already extractive, and workforce dynamics that were already asymmetric. We are accelerating the dysfunction. We are calling it transformation. And the people on the other side of these decisions — the ones we have not asked, the ones we have not consulted, the ones whose livelihoods and identities are being reshaped by choices they did not make — are absorbing the cost.

This is not a technology problem. It never was. It is a transformation problem with a technology component, and the technology is moving faster than the transformation work the system was built to do.

You don’t have to look far to see this happening in real time. In April, Jack Dorsey and Sequoia’s Roelof Botha published an essay arguing that AI can now replace the coordination work middle managers have done for centuries — five weeks after Block cut roughly 4,000 jobs. Their proposed metric for whether the new model is working: money. “The most honest signal in the world,” they called it. Not the workforce. Not the people whose roles were routed away. The money.

That’s not a hypothetical. That’s the pattern, in public, argued in three thousand words, by two people with the platform to set the terms other companies follow.

Subscribe now

I want to bring my parents back into this piece for a paragraph, because I’ve been thinking about them every week of the series.

My mom is a now-retired nurse. My dad was a farmer. In my forthcoming book, I call the chapter about them Soil & Souls, because it’s just the truest way to say it. She tended to souls. He tended to soil. Sacred labor, to both.

Both spent careers developing expertise the new systems are now eroding. My mom learned telehealth mid-career — a screen where a hand used to be, a webcam where her eyes used to meet a patient’s. She adapted. She still doesn’t fully trust a system that can’t feel a pulse change under its palm. My dad watched his instruments shift from the sky and the smell of a field after rain to a sensor that could tell him the moisture level of one patch of ground to the decimal. He would have used every bit of it. He also would have kept walking the field, because that’s where the meaning was, not the dashboard.

Both adapted; both grieved; neither one was asked. Craft or computer was never really the question. Who decides is.

Your workforce is full of my mother and my father. The senior practitioner in your organization, whose tacit knowledge is being codified into a model and deployed to a chatbot. The mid-career professional, whose judgment is being supplanted by a tool that cannot defend its reasoning. The new graduate, whose first job no longer exists because entry level work has been automated. The leader, whose own AI literacy is barely ahead of her workforce’s, and who has not been given the space to learn out loud.

None of these people were asked. Not permission, exactly. Not for a vote or veto. Just for the kind of lived expertise that might have made the decision better, the kind only the people closest to the consequences can give. Instead, they are absorbing a transformation that was decided somewhere above them, on a timeline that did not account for their absorption capacity, with a vocabulary stripped of the words they would need to surface what is happening to them.

That’s the part I cannot let go of. The asymmetry. The decisions being made far from the people who will live with them. The silence the workforce has been carrying because the conditions for speaking were never built.

You are not going to fix this with a better deck. You are going to fix this by interrupting it.

So here is the question I asked you to carry since the start of the series.

What would you do differently if you actually believed the people inside these decisions deserved co-authorship?

I want you to answer this. Out loud or to yourself, but actually answer it. Not in the abstract. About one decision you are making this week. The AI vendor evaluation you are about to sign off on. The pilot scope you are about to approve. The training program you are about to launch. The headcount target someone above you has set without consulting the workforce it will reshape.

Pick one. Hold it in your mind. And ask yourself, honestly: if the people who will live with this decision were in the room when you made it, what would you do differently?

Most readers, I suspect, will arrive at one of three answers.

The first answer is nothing. The decision is fine. The people would have agreed. The process was sound. If this is your honest answer, the decision was probably good. Move on.

The second answer is I would slow down. I would ask whether we have the strategy clear enough to defend. I would ask whether we have done the mindset and heartset work the deployment requires. I would ask whether we have actually named the substitution-versus-augmentation choice or whether we are letting the vendor scope it for us. If this is your honest answer, the decision is not fine. Slow it down. The cost of slowing it down is far smaller than the cost of being wrong.

The third answer is I would not make this decision at all without the people in the room. If this is your honest answer, you have just identified the work. Not the work of redesigning the decision. The work of redesigning the room in which decisions like this get made. That is bigger work. It is also the only work that actually addresses what we have been talking about for seven weeks.

I watched a version of this happen in miniature when my daughter Bailey was 9. She went in for a routine cavity fill, already dreading that alone. Then the plan changed mid-chair — the dentist looked closer, found more than she’d expected, and escalated us to The New Treatment Plan, a root canal, and started walking me through the new estimate. Over Bailey’s head. While she sat small and smaller in the chair, lip quivering, chest heaving under her softball hoodie.

Suddenly, a small voice cut into our big adult ones: “Mom, can I talk to you?”

Everyone stopped dead in their tracks. “Would you like us to leave the room?”

Yes, she nodded slowly.

She didn’t want comfort. She wanted information. Well, both. What is a root canal? Why did the plan change? Is it going to hurt? Do we need to do this? Do you get to stay? With Bailey’s permission, we asked the dentist back into the room, and I asked her to explain it all to Bailey instead of to me. She did, exquisitely — the tools, the steps, what Bailey would feel and when. Bailey listened, and then she told them to turn on the massage chair and go.

Nobody decided anything for her. They decided it with her, the moment she had what she needed to decide for herself. How did she know to do that? I’ve taught her, all three of my daughters, to reclaim her voice the moment she feels it being taken. But knowing how to ask is only half of it. Someone still has to stop and listen.

This time, the room did.

That’s the entire distance between the second answer and the third. My daughter did that work, at nine, in about ninety seconds. Most leadership teams won’t do it in a quarter.

You may have a different answer. The point is not which answer you arrive at. The point is that you arrive at one, honestly, about a specific decision, this week. Most of the time, in most organizations, the question never gets asked. Asking it once is more than most leadership teams will manage all year.

Subscribe now

So here’s the Titanic Test. Before a decision earns its place in your roadmap, it needs three yeses.

Does it expand human potential? Does this move the people it touches toward more capability, or just more automation? If it’s the second, that’s not automatically a no, but name it. Don’t let it pass silently as the first.

Does it create measurable business impact? Can you name the metric this decision should move? If you can’t, it’s just motion, not progress.

Does it strengthen the institution you’re building? Were the people who will live with this decision in the room, or did it arrive fully formed, for them to accept? If the decision could be lifted from any company and dropped into yours without accounting for who works there, how work happens, or what people stand to lose, it isn’t transformation. It’s just boilerplate.

Miss any of the three, and that’s not a yellow flag. That’s the iceberg. You saw it. Now what do you do about it.

That’s the test. Here’s what you do with what it shows you.

A few practices to take with you. Not a checklist or framework. Three things that interrupt the pattern at the layer where the pattern lives.

Notice the silences. In every room you are in this week, watch for the question nobody asks. The participant who almost speaks and then doesn’t. The agenda item that gets deferred for the third time. The follow-up that never happens. Silences in enterprise conversations are not empty. They are full of the thing nobody wants to name. Notice them. You do not have to solve them. Noticing them is the first move.

Make one decision visible that you would normally bury. The Hands, Heads, Hearts framework is not abstract. It is the place where you can name, out loud, what AI is doing to a specific role on your team. Pick one role. Pick one project. Pick one workflow. State, explicitly: we are automating this. We are augmenting this. We are amplifying this. Communicate the decision to the people it affects. You do not have to be certain. You have to be transparent. The transparency is what produces trust. The decision being made invisibly or not at all is what kills it.

Slow one thing down. One pilot. One deployment. One training rollout. Identify it, and slow it by a quarter. Use the time for the mindset and heartset work the standard sequence is rushing past. Most enterprise AI failures are not from moving slowly. They are from moving fast in the wrong sequence. Slowing one thing down to do it correctly is the highest-leverage move available to most leaders this year.

Three practices. None require new budget. None require new vendors. All require the willingness to interrupt yourself, in a meeting, in a decision, in a room where the pattern is currently running.

That is the work.

Seven weeks ago I opened this series with the Titanic in plain sight. The metaphor has been doing the work I needed it to do. The ship was built by people who knew the lifeboats were inadequate, the icebergs were known, and the speed records mattered more than the safety margins. None of those decisions felt wrong in the rooms where they were made. All of them were wrong.

What we’re building now is bigger than a ship. And the rooms where we’re building it are smaller. The decisions are being made by people with little downstream line of sight or care. The icebergs are known, and the lifeboats are inadequate. The speed feels necessary. The speed is the problem.

You are not going to stop this from the outside. None of us are.

We are going to stop it from inside the rooms where the decisions are being made. By interrupting. By noticing. By naming what is happening to people and refusing to let the silence stand for consent. By making the substitution-versus-augmentation choice visible. By building the conditions for the workforce to speak. By letting belief actually constrain behavior. By doing the embodiment work the standard sequence rushes past. By asking the question, every quarter, in every room: what would we do differently if we actually believed the people inside this decision deserved to participate in it?

Noticing is the first act of civil disobedience.

The next act is yours.

Subscribe now

Thank you for reading, all the way from the Titanic in plain sight, or from wherever you found this. That’s not nothing, in an environment built to spend your attention on things that don’t deserve it.

This series ends here. The work doesn’t.

I’ll see you out there.

— JVB

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother, will be published in the fall of 2026.

This is the seventh and final piece in the series. Read them all:

Piece 1:
The Titanic in Plain Sight

Piece 2:
When the Prediction Becomes the Verdict

Piece 3:
The Five Wounds of Change

Piece 4:
It’s Not ‘Will AI Take Your Job.’ It’s Which Choice Your Company Already Made.

Piece 5:
The View From Inside the Room

Piece 6:
A Different Sequence

Piece 7:
What You Choose to Interrupt

The case is built. The wounds are named. The verdict is in.

This week is different. This week I am going to offer you something to do.

What I am about to walk through is the operating model my co-founder Jason Averbook and I have been building at Now to Next, studied across hundreds of transformations, over years of watching what works and what doesn’t. It’s not new. It wasn’t invented. We didn’t sit in a conference room and brainstorm it. We watched it. We watched the same gaps appear in engagement after engagement, in industry after industry, at every scale — and we designed a sequence to fill the gaps we kept finding.

The sequence is called 4SET. The first time you read it, it will look obvious. The second time, you might note it’s in the opposite order from how most organizations are running it. The third time — usually after you have tried to apply it inside your own organization — you will realize that the order is the entire point, and that getting the order right is harder than it looks.

Let me show you.

Subscribe now

Most organizations start at the toolset and work backward.

Here is what happens in nearly every enterprise AI initiative right now. The board approves an AI strategy. The CIO scopes a tool stack. The CTO leads vendor evaluation. The CHRO is asked to build an enablement workstream. The Chief AI Officer — if there is one — translates the strategy into pilots. Training programs roll out. Champions are appointed. Adoption metrics get tracked. Quarterly business reviews celebrate the percentage of employees who have logged into the new tools at least once.

Somewhere down the list, in the appendix slide of someone’s deck, there’s a reference to culture or mindset or change leadership. It rarely has a budget attached. It rarely has an owner.

That’s not an accident. Ownership defaults to whoever already controls the budget line — usually IT, or procurement, or whichever function owns the infrastructure spend. It is rarely whoever should own the transformation. This isn’t a strategy failure. It’s a default nobody decided on, and it’s why toolset ends up first even when everyone in the room agrees it should be last. It’s the workstream everyone agrees is important and nobody actually does, because the operational pressure is on the tools and the tools are where the visible activity lives.

This is the standard sequence. Toolset first. Skillset second (training people on the tools). Heartset third, sometimes, in the form of engagement surveys and town halls. Mindset last, if at all, usually as a reactive measure when adoption has stalled and someone needs to figure out why.

The standard sequence produces standard results. License utilization in the low double digits. Pilot pile-up without scale-up. The ROI gap your CEO can feel but cannot name. The Five Wounds your workforce is carrying but cannot articulate. The 3.4 readiness score we walked through last week.

The standard sequence doesn’t fail because the people running it are incompetent. It fails because it is structurally backwards. You are trying to install behavior before you have built the belief that produces the behavior. You are training people on tools they do not yet trust. You are asking them to embody a future they have not yet been invited to imagine. You are demanding adoption metrics from a workforce that has not been given the conditions for embodiment.

The standard sequence is what we keep telling our clients to interrupt. Here is what we tell them to do instead.

The 4SET sequence.

Four words. They look simple. The order is everything.

Mindset. Heartset. Skillset. Toolset.

We start at mindset and work forward. The reasoning is structural, not sentimental. Each step in the sequence builds the conditions for the next. Skip a step and the steps that follow do not hold. Run it backward — which is what most organizations are doing right now — and the foundation is never there.

Let me walk you through each step.

Mindset.

This is the layer most enterprises skip entirely. It is also the layer everything else rests on.

Mindset is how you see what is possible with AI, and I’m not talking about another Art of the Possible workshop or prompting party. Those have their place, but this refers to openness, curiosity, agency. Not training. Not enablement. Not a mandatory module in the learning management system. The actual cognitive posture a person brings to the work.

Two people can be looking at the same AI tool: One sees a threat to their job. The other sees an instrument that frees them from work they never liked or gives them a thought partner where they really needed one. Same tool. Same context. Same training. Completely different mindset. The first person will resist adoption no matter how good the enablement is. The second person will run ahead of the rollout. Mindset is what determines which one shows up in your workforce, and it is upstream of every other variable you are trying to manage.

Ask any workforce what it wants first, and the answer is almost never a tool. It’s almost always the same request: help me understand what this actually is, what it means for my work, what I’m allowed to do with it. That’s a mindset request. Enterprises hear it and route it to a training module anyway.

Mindset work is not motivational speaking. It’s not poster slogans. This isn’t the CEO video at the all-hands. This is the slow, deliberate work of helping people examine the beliefs they’re bringing to AI — and giving them the time, space, and permission to update those beliefs.

This is what most organizations cannot stomach. Mindset work doesn’t show up on a dashboard. It doesn’t produce a neat quarterly metric. It doesn’t have a deliverable that can be presented to the board. It is also the single highest-leverage investment in the entire sequence, because every step that follows is multiplied by the mindset that precedes it.

The irony is that most organizations interview for mindset — growth mindset, curiosity, adaptability — and then never come back to it once the hire is made. The work is not unfamiliar. It is unpracticed at scale. In fact, every system we build is designed to measure the outputs of mindset rather than the mindset itself.

Skip mindset and you spend the next three years working twice as hard to produce half the outcomes.

Heartset.

If mindset is how you see what is possible, heartset is what you believe about yourself in an AI-augmented workplace.

Heartset is confidence. Belonging. It’s the answer to the question, do I still matter here? It is the felt sense of safety that allows a person to bring their full capability to the work, to make mistakes without fear of being replaced, to learn out loud without losing standing.

Heartset is where the Five Wounds live. Fear of Replacement is a heartset wound. Loss of Mastery is a heartset wound. Identity Disruption is a heartset wound. The workforce is not going to embody a new way of working while it is bleeding from five places nobody is treating.

In the actual rooms where this work happens, the ratio is almost never balanced. When people are finally given room to say what they’re feeling about AI, most of the time in the room goes to naming the fear, not building the skill. Not because people are fragile. Because the fear has nowhere else to go until someone opens the room for it.

This is also the layer most enterprises confuse with the engagement survey. The engagement survey measures heartset retroactively, at the lowest possible resolution, with questions designed to be answered politely. Heartset work is what produces the conditions that would make those survey scores rise in the first place. The work is direct. It is relational. It happens in conversations, in offsites, in the small moments where a leader names the wound out loud and gives the person carrying it permission to set it down.

And heartset, done right, is not a feelings exercise. The organizations that take it seriously build it the way they’d build any other system: visible guardrails, stated boundaries on what AI will and won’t decide, a clear answer to who’s accountable when it’s wrong. Trust isn’t a mood you improve. It’s infrastructure you design.

Heartset is what the bleeding-edge organizations we described last week were trying to build. The champions networks. The communities of practice. The psychological safety work. They were doing real heartset work and not yet getting to embodiment, because the mindset layer underneath was incomplete and the skillset layer above was being scaled too fast.

The order matters. Mindset opens the door. Heartset walks through it. Without the door open, no amount of relational work lands.

Subscribe now

Skillset.

Now we are in the layer most enterprises actually invest in. This is where the training programs live. This is where the prompt engineering certifications live. This is where the AI literacy curriculum lives.

But skillset, in our sequence, means something more specific than tool training.

Skillset is the judgment, creativity, and relational capabilities that remain distinctly human as AI takes over more of the routine work. It is the ability to verify what AI produces, to challenge it when it is wrong, to know when to use it and when to put it down. It is the discernment to recognize that the polished output is workslop, the courage to flag it, and the skill to do the work properly without becoming dependent on the tool.

The most underrated skill in an AI-augmented workplace is the ability to think without the AI. Not as a refusal of the technology. As the underlying capability that lets the technology actually amplify your work instead of dissolving it. The MIT cognitive debt finding I cited two weeks ago is what happens when skillset gets confused with toolset. People become operationally fluent and cognitively atrophied. The skillset layer is what prevents that.

Vivienne Ming’s recent research on human-AI collaboration gives this a name and a number. She ran an experiment putting people through real prediction tasks with AI, EEG monitors attached, and found three distinct patterns. Most people handed the problem to the AI outright — their brain activity on the task dropped to something closer to watching television than working. A second group used the AI only to confirm what they already believed, and performed worse than the AI operating alone. A third group, a small one, stayed in genuine back-and-forth with the tool: pushing, questioning, refusing the easy answer. That group beat the best humans working alone and the best AI working alone, and it barely mattered which AI model they used. What predicted the outcome was the human, not the tool.

Ming calls that third group cyborgs. We’d call it skillset. Same finding, different vocabulary.

Skillset work is also where Hands, Heads, and Hearts come into play — and I want to spend a beat here because this is the second framework that lives inside the sequence.

A second framework, briefly. Hands, Heads, Hearts.

The 4SET sequence is one framework. There is a second one that lives alongside it, doing different work on a different object.

4SET operates on capability. What you build in your workforce, in what order. Mindset, heartset, skillset, toolset. The sequence is about the people doing the work.

Hands, Heads, Hearts operates on the work itself. What AI does — and does not do — across three layers of every job. The framework is about how work gets designed, not about how the workforce gets built.

The architecture is three verbs mapped to three layers.

Hands — automate execution. This is the layer AI is replacing. The routine work, the administrative work, the documentation work. The tasks that used to take hours and can now take seconds. AI is not assisting here. AI is doing the work. The hands of the workforce are being freed for something else, and the question of what that something else is is the entire point of getting the framework right.

Heads — augment thinking. This is the layer AI is changing without replacing. Analysis. Insights. Decisions. AI is not thinking for humans here. AI is making humans better at thinking — faster pattern recognition, broader information synthesis, more rigorous testing of conclusions. The heads of the workforce are still doing the work. The work is just better because the AI is in the loop. (I refuse to say humans in the loop. So punk, I know.)

Hearts — amplify purpose. This is the layer AI cannot touch. Trust. Judgment. Meaning. Connection. The work humans do for reasons AI does not have. The hearts of the workforce are not being automated and not being augmented. They are being amplified — because as routine work compresses and analytical work accelerates, the relative weight of the work that requires a human heart goes up, not down.

This is where trust goes to die.

The substitution-versus-augmentation choice from three weeks ago — the Turing Trap, the headcount KPI, the Dorsey path versus the Palsule path — does not actually get made in a strategy deck or a board meeting. It gets made here, at the Hands, Heads, Hearts layer, in a thousand small decisions about what AI does to which part of which job. Are we automating this hands work, or augmenting the worker who does it? Are we augmenting the thinking at the heads layer, or replacing it with model output and calling it efficiency? Are we amplifying the hearts work, or hollowing it out by removing the relational labor underneath?

Most enterprises never make these decisions explicitly. They buy tools, deploy them, and let the substitution-or-augmentation outcome emerge from whatever the vendor scoped. The aggregate of those un-made decisions is the substitution choice the entire workforce ends up living with, made by no one in particular, communicated through silence, and absorbed by the people who carry the wounds.

This is the mechanism behind every trust failure we are seeing inside enterprise AI rollouts right now. Lack of trust is really a lack of transparency. Lack of transparency is really a lack of decision-making at this layer. The workforce’s trust does not collapse because leadership lied. It collapses because leadership never decided, and the undecided choice was communicated as silence, and the silence was read as the worst-case interpretation. Trust dies in the absence of a stated intent.

The Hands, Heads, Hearts framework is the place where the decision can finally be made visible. Making it visible is what produces trust. Refusing to make it is what kills trust before the rollout even starts.

Automate. Augment. Amplify. Execution. Thinking. Purpose. Three layers of every job, three different relationships with AI, three different design questions for the people redesigning the work.

The two frameworks run in parallel. 4SET builds the capability of your workforce in the right sequence. Hands, Heads, Hearts redesigns the work itself across the three layers and the three relationships. Most enterprises do neither well. They invest in toolset capability while automating hands-only work, and call it transformation.

Back to the sequence.

Toolset.

Most enterprises start here. We end here.

Toolset is how effectively you design and deploy AI tools with reimagined workflows. It is the platform decisions, the vendor evaluations, the integration architecture, the use case selection, the deployment timeline. It is the layer most of the enterprise AI conversation is currently happening in. It is also, in our sequence, the last layer to do real work in — not because it doesn’t matter, but because it cannot do its work without the three layers underneath.

A toolset deployment without mindset produces resistance. A toolset deployment without heartset produces wounds. A toolset deployment without skillset produces workslop. A toolset deployment without all three — which is what most organizations are doing right now — produces the ROI gap your CEO can feel but cannot name.

When the three foundation layers are in place, toolset becomes the easiest part of the work. The workforce knows what they are for. They have the cognitive and emotional capacity to absorb new tools without anxiety. They have the judgment to evaluate the tools critically and the courage to push back on the ones that are not helping. The tools land on prepared ground. Adoption is not a metric anyone needs to track, because adoption is what happens when the rest of the conditions are right.

This is the inversion most enterprises resist. They want to start at toolset because that’s where the budget is, that is where the visible activity lives, and that is what the board is asking about. The longer they start there, the longer they spend trying to fix the consequences of having started in the wrong place.

Why the order matters.

I want to spend a paragraph on why each step has to precede the next, because the order is the entire claim.

Mindset has to come before heartset because what you believe about yourself depends on what you believe is possible. A workforce that thinks AI is going to replace them cannot feel safe inside the rollout, no matter how much heartset work you do. The mindset has to update first. Otherwise heartset is hopeless work on top of an unstable foundation.

Heartset has to come before skillset because skill-building requires risk-taking, and risk-taking requires safety. If a worker is afraid of being replaced, she is not going to expose her gaps to learn new skills. She is going to perform competence she does not have, hide what she does not know, and protect her standing rather than expand her capability. The heartset has to be there before the skillset can grow.

Skillset has to come before toolset because tool deployment without skill produces workslop and cognitive debt. If the workforce has not built the discernment to evaluate AI outputs critically, they will either over-trust the tools and produce poor work, or under-trust the tools and revert to manual workarounds. Either way the toolset investment is wasted. The skill has to be there before the tool can do its work.

This is the sequence. Each step is the precondition for the next. Run it in this order and the work compounds. Run it backward — toolset, skillset, heartset, mindset — and each step gets harder rather than easier, because you are trying to build the lower steps on top of a structure that is already shaped wrong.

Diagnostic before prescription.

One more piece, and then I will hand off.

The 4SET sequence is not a checklist. It is not a methodology to be applied uniformly across every organization. We do not show up at a client and say here is the sequence, run it. We show up and assess where the organization actually is.

Most organizations are not at zero. They have done some mindset work, somewhere. They have built some heartset infrastructure, somewhere. They have some skillset programs running. They have a lot of toolset activity. The work is not to start the sequence from scratch. The work is to identify where the gaps are, in what order, and to sequence the next investments accordingly.

This is what we mean when we say we diagnose before we prescribe. Always. Every engagement begins with a baseline assessment of where the organization is across all four layers, across Hands, Heads, and Hearts. We co-articulate the KPIs the business needs. We do not arrive with predetermined endpoints. Starting points and ending points depend entirely on the organization’s actual journey.

Real implementation doesn’t run the sequence once for the whole enterprise. It runs in parallel, one initiative at a time, each cycling through its own mindset, heartset, skillset, and toolset around a single outcome. Nobody gets to skip the diagnostic because another team already ran theirs.

This is how we get from Now to Next.

The reason this matters is that the standard consulting model produces prescriptions before diagnoses. The vendor sells the tool. The consultant produces the deliverable. Neither one has assessed whether the organization is ready for what is being prescribed. The result is a hundred billion dollars of enterprise AI investment producing the 95% pilot failure rate everyone is now reading about.

You cannot fix that with better prescriptions. You can only fix it by starting with better diagnoses. The 4SET sequence is the diagnostic frame. Hands, Heads, and Hearts is the dimensional frame. Together they give you a way to look at any AI deployment in your organization and identify where it is going to fail before it does.

What this means for you.

You have a choice this week. You can keep running the standard sequence — toolset-first, with mindset as an afterthought — and accept the standard results. Or you can interrupt the pattern.

Interrupting the pattern does not require ripping up your current AI strategy and starting over. It requires three things.

One. Sequence audit. Look at the AI investment you have already made and ask yourself, honestly, what layer you started in. If you started at toolset, the next investments need to fill in the foundation, not extend the tooling. If you started at skillset with no heartset work, the next investments need to address the wounds. If you have done foundation work but it has been disconnected from the tooling layer, the work is to connect them.

Two. Dimensional check. Whatever step of the sequence you are working in right now, ask whether you are doing the work in Hands, Heads, and Hearts simultaneously. If you are only redesigning the operational layer, you are leaving two-thirds of the value on the table. If you are working on culture without redesigning the work itself, you are producing inspiration without infrastructure.

Three. Diagnostic discipline. Stop prescribing before you have diagnosed. Every initiative you are about to launch should begin with an honest assessment of where the workforce, the function, and the organization actually are. The diagnostic does not have to be elaborate. It does have to be honest. The reason the standard sequence produces standard results is that nobody pauses long enough to ask where they actually are before deciding where to go next.

These three moves do not require new budget. They do not require new vendors. They do not require new technology. They require new vocabulary, new measurement, and the willingness to slow down for the diagnostic work that the standard sequence is rushing past.

The organizations that do this are already pulling away. Not loudly, not visibly to the market yet. But the ones doing the embodiment work, in the right sequence, across the right dimensions, are producing outcomes the standard sequence cannot. They are the early signal of what comes next.

Next week, the close. The seventh piece in this series. The one where I will not introduce a witness, will not name a framework, will not surface new evidence. The one where I will ask you a single question and ask you to answer it out loud, or at least to yourself.

The question is the one that has been running underneath every piece in this series. You already know what it is.

The piece is called What You Choose to Interrupt.

I’ll see you next week.

Pictured: Brandenburg Gate in Berlin last week. Loaded history put simply: it began as a symbol of power, became a symbol of division, and survived to become a symbol of unity. Punk Berlin, I loved you.

Subscribe now

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother, will be published in the fall of 2026.

This is the sixth in a seven-part series. Read Piece 1, Piece 2, Piece 3, Piece 4, and Piece 5 here. On the home stretch now.

This week the witness is the work itself. The rooms we’ve been in. What we hear inside real conversations in all the muck and glory of real initiatives. Two recent engagements, anonymized as conceptual case studies, both highly representative of patterns we see nearly everywhere we work. Read what follows the way you would read a portrait, please. Not a study. Not a deck.

Engagement A.
A global enterprise People function.

We ran an engagement with the senior People and Culture leadership of a global enterprise — People Officers, Heads of Talent, strategy owners running innovation programs across business units. The CHRO stood at the front of a very senior room and named what most CHROs don’t: that the People function had to lead this transformation or it would be done to them. The room agreed. The energy was real. People left clear, committed, ready.

The harder question was what would survive the week.

So we asked them. Not with a survey. We’ve been prototyping a different kind of listening practice inside Now to Next — an AI-powered voice interviewer participants can engage on their own time, by phone, anywhere, anonymously. They speak instead of typing. They wander (and wonder) instead of clicking. They reflect after the event has landed, when they’re not in a meeting, not in front of their team, not performing for anyone. The questions are open. The answers are theirs. Nobody who reads the findings can identify who said what. The format itself is an act of psychological design — we wanted to know what they would say if they were certain no one would know they had said it.

What came back is the closest portrait we can offer of the Five Wounds happening to real senior leaders in real time. Four patterns worth walking through.

Pattern one:
The energy was real. The operating reality was not.

Most of the senior leaders said some version of this directly. They left the session committed, then they went back to their desks. They hit the same legacy systems, the same approval cycles, the same policy gates that had been in place before. I still felt energized, but we have so many roadblocks.

This is the gap every transformation initiative has lived in. AI is widening it rather than closing it, because investment is being concentrated on the technology layer while the operating-reality layer is being neglected. Trust Erosion lives here. Every time the room is energized and the operating reality is not, the workforce learns that the room is theater.

One leader observed, almost as an aside, that she had been surprised by how often the word hope came up in the session. The fact that hope appearing in a business context registered as unusual is the data point. We have built an enterprise discourse so dominated by urgency and threat that hope has dropped out of the vocabulary entirely.

Pattern two:
People using AI every day called themselves beginners.

The leaders we spoke with use Claude or Copilot or Gemini daily. Many of them for over a year, but they still called themselves beginners. What they meant, when pressed, was that they wanted AI embedded into the workflows they already had — not a separate destination they had to remember to visit.

This is not a technology problem. This is a design problem. The tools have not been adapted to the work. The work has been forced to adapt to the tools. Even sophisticated users feel like beginners — not because they are, but because using AI in their actual job is still effortful, still kludgy, still requiring active accommodation. Decision Fatigue lives here. Not from the AI. From the constant context-switching the current generation of tools demands.

The function knows what it needs. When we asked senior leaders what their function had to do exceptionally well in the next three years, the words filling the room were transparency, communication, evolve, innovate, change management, deliver value, enable business strategy. Mostly the function’s relationship to the business and to the workforce. Not technology. Not platforms. The People function is naming its own job correctly. The question is whether the rest of the organization is going to let them do it.

Pattern three:
One leader named the question nobody else did.

I quoted her in Piece 3 without identifying her. The substance: how do we tell people the truth about what AI will do to their jobs, in a way that brings the workforce along rather than losing them?

The data point is not that one in ten leaders ask this question. The data point is that one in ten name the question everyone has, while the rest carry it privately. The wound is present in everyone. The willingness to surface it is present in one.

The easy interpretation is wrong. The ones who do not raise the question are not less courageous, nor less aware. Even when the conditions for saying it are carefully built — anonymity, voice, off-hours, no name attached — those conditions are not safe enough. That tells you something about how much weight the silence is carrying. The senior leaders charged with leading AI transformation are themselves carrying the question they cannot ask their workforce. They are not above the wound. They are inside it.

Pattern four:
Time was the scarce resource. And the reason it was scarce was strategic.

When we asked senior leaders what was holding them back, they didn’t say they needed more training or better tools. They said they did not have the hours. They were performing full-time jobs that had not yet been modernized using the tools they were supposed to be deploying, being asked to lead the AI transformation on top of the existing work, with no relief on the existing work.

When we asked them to name the biggest barrier between today’s People function and the future state they had just described, the answer ordering was instructive. Largest barrier: lack of strategy. Second: skillsets and education. Third: fear and uncertainty. Smallest: tool access. The People function is telling you, unprompted, that the technology is not the constraint. Strategy is. Confidence is. Tools are barely on the list.

These two findings are not in tension. They are the same finding seen from two altitudes. The capacity crisis the leaders are living through is what the strategy gap feels like from inside the work. When there is no clear strategy from above, every senior leader is constructing the work from scratch — every meeting, every decision, every prioritization call. That construction is what consumes the hours. Tell a senior leader the strategy clearly, and you give her back the hours. Leave her to figure it out by inference and reverse-engineering, and you bleed her capacity at exactly the rate the transformation demands more of it.

When we asked how ready the People function was to make the shift it had just defined for itself, the average was 3.4 on a 5-point scale where the right end was ready. Just above the midpoint. Not paralyzed. Not confident. Aware of the gap.

Stop on that number. A senior People function inside a global enterprise — the executives charged with leading their workforce through the largest transformation of their careers — rated their own readiness at 3.4 out of 5. Not in a performance review, where the number would have been higher. Not in a town hall, where the number would have been higher. In an anonymous reflection, by voice, on their own time, knowing nobody would attach their name to the answer. The honest number is just above the midpoint. The function knows what is required. It is rating itself at 3.4 on whether it is ready to deliver it.

This is the structural mistake the series has been indicting. We are running an enablement program where we need a transformation. We are scaling tool adoption when we need to scale human capacity. The capacity is finite. The clock is not extending. And the function being asked to lead the work is rating its own readiness at 3.4.

Engagement B.
The ROI gap the client could feel, not name.

In another engagement, we sat with the leadership team at a high-growth technology company that had been running AI initiatives for two years. By any conventional measure they had been doing this well. They had identified champions, stood up an internal community, carved out time for learning and for sharing. They had given people psychological permission to experiment, and built the structures to capture what came out of the experimentation. Adoption metrics were good. Cultural posture was healthy. They’ve honestly done the work most enterprises are still trying to figure out how to do.

And yet. An executive sponsor said some version of we have been doing this for two years and we still don’t… — and stopped. On a cliffhanger.

If pressed, he might explain how the technology had changed and the tools had changed and the processes had changed and the dashboards had changed, but the actual work of the actual workforce had not changed. The relationship between humans and the systems they used had not changed. The way people thought about their own roles and capabilities had not changed. They were getting all of the adoption and none of the transformation. It was registering as a tightness no one could quite locate, so they asked us for help.

This is the ROI gap almost every enterprise leader is feeling right now, even the ones doing the adoption work well. The gap is not because the adoption work is bad. The gap is because adoption and transformation are different categories of work, and the field has been conflating them for two years.

Jason Averbook and I have been making this argument to every client we work with: Technology is not transformation. Adoption is not embodiment. Adoption is whether people use the tool. Embodiment is whether they think, collaborate, and solve problems differently because of it. The first is a usage metric. The second is a way of being. Most enterprises are measuring the first and hoping it produces the second. It does not. It cannot.

What was unique about this client was not that they had figured this out themselves. They had not. What was unique was that they could feel the gap. They had done all the work the field tells you to do, and they could feel that something was still missing. The unfinished sentence was the recognition. That recognition gave us the opening to introduce the Five Wounds. Once the wounds were named, the team could see where the existing adoption work was succeeding and where it was failing.

The recognition is the most valuable thing a senior leader can bring to this work. Not the language. We have the language. The recognition that something is missing here that our current playbook cannot reach is the door opening.

What both engagements share.

A note before I tell you. Engagement A and Engagement B look bleeding edge at face value. They have invested heavily — not just in AI tools, but in adoption programs, internal communities of practice, executive sponsorship, offsites for learning and psychological safety. By any conventional measure of enterprise AI readiness, they are ahead of their peers.

That is part of why we’re sharing them. What shows up below is what happens when you have done the foundational work and are still not getting to the outcomes you can feel are possible. The advice in next week’s piece is sequenced for organizations that have already invested the way these two have. If you are still building the foundation, you have a different set of investments to make alongside the sequencing recommendations. The sequencing doesn’t skip the foundation. It assumes you have built one, or are building one in parallel.

These two engagements are the most illustrative pattern we are seeing right now. They are not the only pattern. We expect other shapes to emerge as more organizations push further into this work. What we can tell you is that the pattern we see in Engagements A and B is showing up in nearly every advanced engagement we run.

Engagement B proved something important: adoption isn’t transformation. They had done everything the playbook prescribed—deployed the tools, activated champions, built the community, delivered the training. Adoption was high. But people weren’t thinking differently, collaborating differently, or solving new kinds of problems.

Meanwhile, Engagement A never made it that far because it lacked the capacity to begin.

Different organizations. Same bottleneck.

One couldn’t start the work. The other finished the wrong work.

The problem isn’t execution. It’s sequence. Most AI transformations still follow a digital-era playbook: deploy technology, train users, manage change, and hope new behaviors emerge. But embodiment isn’t the byproduct of adoption. It has to be designed from the beginning.

AI is not a software rollout. AI is a cognitive and cultural shift with a software component, and the sequence that worked for software is the sequence that is failing for AI.

Next week, I’ll introduce the framework we use to design AI transformation differently: mindset, heartset, skillset, and toolset.

It’s not a linear checklist, and organizations rarely move through it in a neat sequence. In practice, the work overlaps, loops back on itself, and unfolds unevenly. But ignore any one of these dimensions, and the gap shows up somewhere else. Every shortcut creates a wound. Every missing layer becomes tomorrow’s resistance.

The piece is called A Different Sequence.

The question carries.

What would you do differently if you actually believed the people inside these decisions deserved co-authorship?

I’ll see you next week.

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother, will be published in the fall of 2026.

This is the fifth in a seven-part series. Read Piece 1, Piece 2, Piece 3, and Piece 4 here.

Among young workers ages 22-25, in the occupations most exposed to AI, employment has declined sixteen percent since late 2022. The decline is concentrated in roles where AI is automating the work rather than augmenting the worker. The decline holds when researchers control for interest rates, for the pandemic, for remote work, for the tech sector. The decline is not a recession effect. It is not a cohort effect. It is not a hiring freeze effect that will resolve when the economy rebounds.

The decline is the consequence of a choice your organization is making right now, and almost no one in your leadership team has named that choice out loud.

This is the third witness in the series. Last week, the Amodeis testified to what it costs when an organization’s stated values do real work. This week, Erik Brynjolfsson is going to testify to what it costs when stated values are absent and the financial logic is allowed to run unchallenged. He has the data. He has been collecting it for three years. He is the witness whose testimony you cannot dismiss as soft, alarmist, or partisan, because he is none of those things. He is an economist at Stanford, the director of the Stanford Digital Economy Lab, and the closest thing the AI-and-labor conversation has to a referee.

What the referee is now saying is that the game is going badly, and the people losing are not who you think.

Erik Brynjolfsson is a useful witness for one specific reason. He has spent his entire career — including a long arc at MIT before Stanford — making the case that technology can be a positive force for workers. He’s not a doomsayer. He is on record arguing that AI could lead to less income inequality, not more. He has been the rare voice telling enterprise leaders that the productivity gains are real and worth pursuing.

That is why his current findings carry the weight they do. When a believer changes his mind even partially, it counts more than a hundred skeptics confirming what they already thought.

Brynjolfsson has not changed his mind on the productivity gains. He still thinks they are real. He thinks they will compound. He thinks the United States will be measurably more productive in five years than it is today.

What he has changed his mind on is the question of who benefits, and who gets hurt in the transition, and whether the choices being made inside enterprises right now are aligned with the version of the future he originally argued for. The answer, increasingly, is no. And the data is starting to show it.

The Turing Trap.

This is the framework that holds everything else in the piece together, and it is worth slowing down for.

Alan Turing, in 1950, proposed that the test of artificial intelligence was whether a machine could imitate a human well enough that another human could not tell the difference. The Turing Test, as it came to be known, set the agenda for the next seventy-five years of AI research. Build machines that act like humans. That was the goal.

Brynjolfsson argues this was the wrong goal. AGI has become a synonym for human-like intelligence or superhuman intelligence, he says, and I can see why that’s an easy benchmark, but most technologies complement humans. They don’t replace humans. AI is no different. The frontier of useful AI is not in machines that do what humans do. It is in machines that do what machines are good at — pattern recognition at scale, calculation, retrieval, synthesis — in support of humans doing what humans are good at, which is judgment, relationship, interpretation, and decision under genuine uncertainty.

The Turing Trap is what happens when an industry, and the enterprises buying from that industry, mistake the imitation goal for the value goal. The AI companies build models that approximate human reasoning. The enterprises measure success by how many humans the models can replace. The CFO sets the KPI as headcount reduction. The CIO scopes the pilots around substitution. The CHRO is asked to manage the transition.

And nobody asks the question Brynjolfsson keeps pressing on every executive he meets: what is the AI tool I can use to make Bob and Sally maximally efficient? He points to one CIO who pushed back and asked how to get workers to buy into AI more. Brynjolfsson’s answer: Stop telling them you’re going to use it to replace them.

That is the Turing Trap, in one exchange. The technology was designed to imitate humans. The enterprise was designed to replace humans. Nobody chose this consciously. Both choices were made by default. The workforce knows. The workforce has known for two years. And the workforce is responding rationally — to what they have heard, and to what nobody is bothering to deny.

The CFO conversation.

There is one moment from Brynjolfsson that I want every CFO and CSO reading this piece to sit with for a minute.

He was meeting with the CFO of a very large company. She told him the company wanted hard measures of AI performance. The next sentence: we are going to measure how much headcount reduction there is in each division.

Brynjolfsson’s response, paraphrased: Okay, that’s one measure you could have. I get why it’s the easiest to count and you like to count things. But you’re a CFO. Your job is to come up with new KPIs. How is customer satisfaction? How many new products? Which quality? I know it’s harder than counting headcount. But most of the value is on the other side.

I want you to read that exchange as the moment the Turing Trap closed around an enterprise in real time. The CFO is not malicious. She is doing her job. She is being asked to deliver financial returns on a major investment, and the easiest measurable return is the one she defaulted to. Nobody in her leadership team told her this was the wrong measure. Nobody had a better one to offer. She did not invent the headcount-reduction KPI. She inherited it from a hundred prior productivity initiatives where the measure worked well enough to defend in a board meeting.

The problem is that AI is not a prior productivity initiative. AI is the technology that lets the substitution choice be made at a scale and a speed no previous technology permitted. When the CFO of a Fortune 500 company sets headcount reduction as her primary AI KPI, she is committing the organization to substitution. She is not augmenting. She has chosen. And nobody around her table has named the choice as a choice.

Most enterprise AI strategies are currently being run by some version of this CFO. The choice is being made by default, in language no one has translated into the language of consequence. Your workforce is the consequence. They are already feeling it.

Canaries in the Coal Mine.

The sixteen percent number I opened with comes from a paper Brynjolfsson published last year with Bharat Chandar and Ruyu Chen at the Stanford Digital Economy Lab, since updated with a broader control set that sharpened the figure. They titled it Canaries in the Coal Mine, which is exactly the right metaphor: a small number of vulnerable workers showing early warning of a danger the larger system is not yet registering.

The method is careful. Brynjolfsson and his team partnered with ADP, the payroll processor, to get access to tens of millions of payroll records. They ranked every occupation in the US economy by AI exposure — how much of the work in that occupation can be done by current AI models. They tracked employment changes by occupation, by age cohort, by industry, by region.

When they first looked at the top-line data, they almost wrote a paper saying there was nothing yet to find. The aggregate numbers showed noise, not signal.

Then Chandar and Chen looked at subgroups. The pattern emerged. Among workers aged 22-25 — recent college graduates entering the labor market — in the most AI-exposed occupations, employment was down 16% since late 2022. In specific occupations like coding and call center work, the declines were steeper still; software developers in that age band fell nearly 20% from their late-2022 peak. Among older cohorts in the same occupations, the declines were negligible.

The critique came fast: the data starts in 2022, so this can’t be AI, it’s the pandemic, or interest rates, or remote work fallout. Brynjolfsson’s team tested every one of those hypotheses. They controlled for interest rates and the effect held. They controlled for the pandemic and the effect held. They controlled for remote work and the effect held. They controlled for the tech sector specifically — a sector with its own hiring dynamics — and the effect held. In some controls, the effect actually got stronger. Construction, which is highly exposed to interest rates, showed no AI-exposure decline. The two variables moved in opposite directions.

Brynjolfsson now believes the real onset of the AI employment effect started in 2024, not 2022, and that earlier data noise reflected pandemic and economic dynamics that have since resolved. The current trend, he says, is getting steeper. It is statistically significant. And it is concentrated, with surgical precision, in the workers who can least afford to lose their first job: the ones at the beginning of their careers, in the occupations most directly substitutable, with no prior employment history to fall back on.

This is Fear of Replacement, but it is no longer fear. It is the data. The young workers are not losing the jobs they have. They are not getting hired into the jobs they expected. The career ladder is having its bottom rungs quietly removed, and nobody has told the people standing at the base of the ladder that the rungs are gone.

You will not see this in your engagement survey or in your attrition report. You won’t see it in your AI adoption dashboard. You will see it in the slow shift of who is no longer in your hiring funnel — the candidates that used to apply and no longer do, the entry level requisitions that never got opened this year because we can just use AI for that now. The Canaries finding is what happens at the level of the entire economy when every individual enterprise makes the substitution choice independently. The aggregate effect is what nobody chose, and yet here we are.

Brynjolfsson’s team is not the only one seeing it. Revelio Labs, tracking job postings rather than payroll records, finds highly AI-exposed entry-level roles down more than 40% — the steepest decline of any category they measure, worse than low-exposed entry-level roles and far worse than any non-entry-level cohort. Two different data sources, two different methods, one economist and one labor analytics firm who did not coordinate. Same shape.

The rebuttal you will hear — and why it isn’t one.

Two days before this piece went out, a paper from Ramp and Revelio Labs started circulating with the kind of headline that travels fast: firms with the highest AI spending grew total employment by roughly 10% over two years, and entry level hiring at those firms grew 12%. Low-intensity adopters saw no change. The framing on the way out the door, in the coverage and on social media, was blunt — AI isn’t killing jobs, actually. Someone on your leadership team will forward you this study by tomorrow. Have your answer ready before they do.

Here it is. The Ramp/Revelio paper and the Canaries paper are not measuring the same thing, and the difference is the entire argument of this piece.

Canaries asks a question about occupations, across the whole economy: if a job is high in AI-exposed tasks, what happens to employment in that job, everywhere it is performed? The answer is a 16% decline for the youngest workers in it.

Ramp/Revelio asks a question about firms, inside a self-selected sample: among 21,559 companies that chose to spend heavily and consistently on AI, what happened to their total headcount? The answer is 10% growth — and even Revelio’s Chief Economist Lisa Simon will not call it causal. They flag it themselves: the heavy adopters in their sample were already larger, faster-growing, more technical, and more likely to be venture-backed before they spent a single dollar on AI tools. Fast-growing companies buy a lot of software. That correlation runs in a direction you would expect with or without AI in the picture.

Read the two studies side by side and you do not get a contradiction. You get the mechanism. Fast-growing, well-capitalized firms are hiring into the roles where AI augments rather than automates — engineering, sales, administration, customer service — and it shows up in their books as entry-level growth. Meanwhile, across the wider economy, the roles where AI automates rather than augments are the ones shedding the youngest workers first, and those losses land hardest at companies that don’t share that sample’s growth profile or hiring budget — the ones scaling their broken processes with AI instead of redesigning them.

Both things are true at once because they are the same choice, made by different companies, on opposite sides of the substitution-versus-complementarity fork this piece is about. The Ramp/Revelio paper is not evidence against the Turing Trap. It is a photograph of what the other path looks like when a company chooses it deliberately — the path this piece is asking you to choose too.

The call center finding everyone cites — and the follow-up nobody is talking about.

If you have read anything about AI’s effects on the workforce in the last two years, you have probably read about Brynjolfsson’s call center study. It came out in the Quarterly Journal of Economics last year, co-authored with Danielle Li and Lindsey Raymond. The headline was the most cited finding in the AI-and-labor literature: AI assistance produced a 15% average productivity gain in call center workers, with a 30% gain among less experienced and lower-skilled workers, and only small gains — with a small quality decline — among the most experienced agents.

This was the AI as equalizer story. It compressed the productivity gap. It helped the people at the bottom of the skill distribution most. It seemed to confirm Brynjolfsson’s original optimistic thesis: AI as a force for reduced inequality. Every consultant report, every vendor deck, every think piece arguing for the bright future of AI-augmented work has cited this study. It is the empirical anchor of the entire optimistic case.

Here is what Brynjolfsson told Nicholas Thompson in their conversation, almost in passing, about the follow-up paper he is now working on with the same company.

The company is mostly using AI agents to answer the questions directly. Only a few of them get escalated to humans... Then you end up having fewer people.

Read that sentence again. The same company. The same workforce. Two years after the celebrated study that showed AI compressing the productivity gap and lifting the lowest-skilled workers. And now the call center is mostly being handled by AI agents directly, with humans only on escalations. The compression effect was real. It was also a transitional phase. The endpoint, on current trajectory, is fewer humans, doing harder work, with less of the volume needed to keep the workforce employed at its prior scale.

Brynjolfsson did not bury this. He said it directly. But almost no one in the enterprise discourse is talking about it, because it complicates the optimistic story everyone has been telling each other. The egalitarian phase of an AI rollout, in this single most-cited case study, was the temporary phase. The substitution phase came next. And the substitution phase is the steady state.

This is the most important sentence in this piece, so I am going to say it twice:

The egalitarian phase of an AI rollout may be the temporary phase. The substitution phase is the steady state.

Your AI strategy is almost certainly modeled on the egalitarian phase. Your business case assumed it. Your communications to the workforce promised it. Your CHRO is currently celebrating modest adoption gains that mirror the early call-center findings. None of this is wrong. It is also, on current evidence, not durable.

The workforce knows. The workforce has been watching this pattern play out across every prior wave of automation. They know what comes next. The reason they are not engaging with your AI rollout with the enthusiasm you expected is not that they do not understand the tools. It is that they understand the pattern, and they are watching for the moment the augmentation language becomes the substitution language. In many companies, that moment has already come. In others, it’s six quarters away. None of your workers believes it will not come.

This is Loss of Mastery and Identity Disruption arriving in the same wave. The experienced practitioners watch their tacit knowledge get codified into the model. They see what their juniors see — that the codified version is increasingly being deployed without them in the room. They lose mastery to the codification. They lose identity to the substitution. And they have no language to say what is happening, because the official language of the AI rollout is still augmentation.

The four moves Brynjolfsson recommends.

He has been asked, by every interviewer and policymaker who has him in a room, what enterprises should actually do. He has a list. I am giving you his list, but with translations into the language of the People & Culture leaders this series is written for.

One. Better measurement. Brynjolfsson and the Stanford Digital Economy Lab are building a set of AI economic indicators — what he half-jokingly calls a shadow AI BLS — that will give organizations near-real-time visibility into AI’s effects on employment, wages, and productivity. The translation for your organization: stop measuring AI deployment by adoption metrics and time saved. Start measuring it by output quality, employee satisfaction, and workforce composition over time. The metrics you currently have are designed for a different question.

Two. More dynamism. The US economy has become less dynamic, not more, despite decades of technology change. Fewer workers move between jobs, companies, regions, or industries than twenty years ago. AI’s transition costs will be borne disproportionately by workers who cannot move. Brynjolfsson recommends portable benefits, retraining programs, and reduced friction in labor mobility. The translation for your organization: your AI strategy needs an internal mobility strategy. The workforce you have today is not the workforce your post-AI operating model needs, and the gap will not close on its own. Most enterprises have outsourced this question to the labor market. The labor market will not solve it on the timeline AI is moving.

Three. Lean into complements, not substitutes. This is the Turing Trap reversal at the operating level. Ask, of every AI deployment: does this make my best people 10x more effective, or does it remove the need for my best people? Brynjolfsson is explicit that he is mad at the AI companies for designing AI to imitate humans rather than complement them. The enterprise version of this anger should be directed at your vendor evaluation criteria. If your AI procurement scorecard rewards substitution capability and ignores complementarity, you are buying yourself into the Turing Trap. The vendors are responding to the demand you are creating.

Four. Centaur benchmarks. The current benchmarks measure how well AI performs alone. They should measure how well AI plus human performs together. Brynjolfsson cites medical imaging: the right benchmark is not how well does the AI recognize cancer in this scan but how well does the radiologist plus the AI recognize cancer in this scan, with an explanation the radiologist can verify. The translation for your organization: your AI tools should be evaluated on how much they improve the work of the humans using them, not on how much they automate the work of the humans they replace. If your vendor cannot show you the centaur benchmark, they have not done the work.

These four moves are not theoretical. They are concrete. They are also, in most enterprises, unaddressed. The CFO with the headcount-reduction KPI is not measuring any of them. The CIO scoping pilots for substitution is not measuring any of them. The CHRO running the change management program is not measuring any of them, because the metrics she has are downstream of the choice that has already been made elsewhere.

The question Brynjolfsson keeps asking.

There is a line Brynjolfsson uses that names the deepest question in the entire AI-and-labor conversation, and it’s the line I want to leave with you for the rest of this week.

He was describing the difference between the optimistic and pessimistic AI futures, and he said this: What is the speed of change that we can absorb as a society, and how do we make society more resilient so we can absorb a higher speed of change?

That question — what speed of change can the system absorb — is the question your organization is also failing to ask, at a much smaller scale. Your workforce has a speed of change it can absorb. Your culture, your operating model, your customer relationships all have a speed of change it can absorb.

You are deploying AI at a speed substantially exceeding all of those limits, and you are calling the resulting friction change fatigue. This is not resistance to change. It is the predictable response of a system being pushed past its absorption capacity.

This is Decision Fatigue at the organizational level. Your workforce is not refusing to adopt. They are running out of bandwidth to absorb. Your senior people are not refusing to mentor. They are running out of capacity to translate. Your culture is not failing to evolve. It is being asked to evolve faster than any culture has ever evolved, with less support than any prior transformation provided, in service of an outcome no one in the workforce was asked to validate.

The speed of change question is the operational version of Daniela Amodei’s we can only diffuse this at the speed of trust. Both are asking the same thing: what does the human side of the equation actually require, and is anyone protecting it?

In most organizations, the answer is no.

Two CEOs who named the choice.

I want to introduce you to two CEOs who did something almost nobody else in enterprise leadership has done. They named the choice out loud, in public, in writing. They are not anonymous. They are not composite, and they are not theoretical. They are running real companies, in 2026, and they made opposite decisions about the same question.

The first is Jack Dorsey, the co-founder of Twitter and CEO of Block. In late February 2026, he announced that Block was cutting 4,000 employees — approximately 40% of its workforce — and tied the decision directly to AI. “We’re already seeing,” he wrote, “that the intelligence tools we’re creating and using, paired with smaller and flatter teams, are enabling a new way of working which fundamentally changes what it means to build and run a company.” In a separate letter to shareholders, he predicted the rest of the industry would follow. “I think most companies are late. Within the next year, I believe the majority of companies will reach the same conclusion and make similar structural changes. I’d rather get there honestly and on our own terms than be forced into it reactively.”

Read that quote again. I’d rather get there honestly and on our own terms. Dorsey did not drift. He chose. He chose with intention, in writing, on the record, with full awareness that the cost was being paid by four thousand of his employees. He did the substitution math and he stood behind it. He framed his approach as bravery — we’re already seeing it, the rest of you are late, let’s stop pretending. Block’s stock went up on the announcement. Investors rewarded the clarity. The 4,000 employees who lost their jobs did not get a vote, but transparency is respect too, even if it comes late and not early.

I want to be precise about something. Some industry observers have argued that Dorsey’s AI framing was partly retrospective — that Block over-hired during the pandemic and that a significant portion of the cuts would have happened regardless of AI. One former Block employee called it organizational bloat wearing an AI costume. That critique deserves to be on the record. But it does not change the broader point: Dorsey publicly chose to brand the cuts as the AI-enabled future of work, and other CEOs heard him, and several have already begun describing their own restructurings in the same language. The signal he sent into the industry was the substitution choice, made boldly and with transparency, framed as competitive necessity. Whether his particular math fully held is less important than the fact that he gave permission. Within weeks, AI governance consultants reported a wave of mandates from boards: every employee must be using AI, every team must show efficiency gains, the substitution clock is ticking.

The second CEO is Himanshu Palsule of Cornerstone, the global talent and learning technology company, recently rebranded as an intelligence platform for workforce readiness. He posted on LinkedIn just weeks ago, and the post is the cleanest counter-anchor to Dorsey I have seen from a CEO at scale.

“The greatest irony of our time?” he wrote. “We built this generation of AI native thinkers and now we’re turning them away at our corporate doors. We handed them an iPad at age five. We bragged at dinner parties when they figured out smart phones before they could ride a bike. We watched proudly as they navigated four different AI tools to finish a homework assignment that would have taken us a full weekend at the library. We literally engineered their minds around technology. Let’s stop calling them ‘Gen Z’ like it’s a diagnosis. We made them. And now, now that the workforce needs people who think that way, we’re rejecting them?”

He closed with the operational commitment: “At Cornerstone, we’ve been doubling down on hiring college graduates — not out of charity, but out of conviction. Because we believe the skills that look like liabilities in a traditional interview are exactly the capabilities this next chapter of work demands. What would it look like if we stopped interviewing for the jobs of yesterday and started hiring for the work of tomorrow?”

Palsule made the opposite choice from Dorsey. He chose complementarity. He is choosing to hire the workers Brynjolfsson’s Canaries data shows are getting locked out of the economy. Which means choosing to read the AI-native cognitive profile of young workers as an asset rather than a liability. He doesn’t frame it as charity or social responsibility; he frames it as competitive advantage. And he did it in public, on the record, with his company’s name attached.

Two CEOs. Same question. Same evidence about where AI is heading. Opposite decisions, publicly committed.

The substitution choice is not a hypothetical you can avoid by staying quiet. It is being made every quarter, in every enterprise, by every leader who has not interrupted the default. The two CEOs above are the rare ones who made the choice visible. I’m not here to judge the choice, I’m here to point out that most leaders are making it without naming it. The naming is the difference between Dorsey’s honesty and your organization’s drift. The naming is also the difference between Palsule’s conviction and your organization’s good intentions.

You will not be in either of these CEOs’ positions exactly. Your scale is different. Your industry is different. Your workforce is different. But you are making the same choice they made. Right now, this quarter. The version of the choice you are making is currently invisible to your workforce and probably to your board. It is visible only to the systems that count and the people who carry the cost.

Witness three, testimony complete.

The choice has been made. The question is whether you are going to name it.

The substitution path was chosen, by default, in the language of efficiency, by leaders measuring what was easiest to count. The Turing Trap closed around your enterprise the moment the CFO set the headcount KPI and nobody offered an alternative. The Canaries finding is the early data on what that choice produces. The call center reversal is the late data on what that choice produces. The speed-of-change question is the warning the system is sending you that the current trajectory exceeds the absorption capacity of the humans inside it. And Dorsey and Palsule are the public bookends of the choice, demonstrating that the two paths are not theoretical. They are operational, and they are being chosen by real CEOs at real scale, in opposite directions, this year.

You have a choice. The question is whether you are going to name the choice, surface the alternative, and design the AI deployment around the complementarity goal rather than the substitution goal. That choice does not require new technology. It does not require new investment. It requires new measurement, new vocabulary, and the willingness to interrupt the financial logic that has been making the choice for you.

Brynjolfsson said one more thing I want you to hold. It is the most generous sentence I have heard him use, and it is the sentence that should anchor your AI deployment conversation for the next year.

We have incredible agency, but we are squandering that because we don’t understand well enough what our choices are.

You have agency. You have not been told that you do. The choice your CFO is making by default can be made differently. The headcount KPI can be replaced. The vendor scorecard can be rewritten. The AI deployment can be designed around the question: how do we make our best people maximally effective instead of the question: how do we replace them. The two questions produce entirely different technology, entirely different change management, entirely different workforce experience, and entirely different economic outcomes.

Most enterprises will drift into the substitution choice. A few — Dorsey is one — will choose it on purpose. A few more — Palsule is one — will choose complementarity on purpose. The drift is what most CHROs and CSOs and CAIOs are doing right now, whether they realize it or not. The drift is reversible. The cost of reversing is the friction of naming the choice out loud, in the rooms where the choice is currently invisible, with the people who don’t yet know that they are the ones being asked to choose.

Or rather, the drift is reversible right up until it isn’t. Ask 4,000 former Block employees.

Next week, witness four. The first witness from inside the enterprise. The first witness I can speak for in my own voice, drawn from two recent engagements where the patterns the prior three witnesses have described are showing up — not in academic papers or interview transcripts, but in conference rooms and post-session interviews with the people doing the work.

What you’ve read so far is the case from outside. Next week is the case from inside. The Five Wounds in the rooms where they are happening. The vocabulary your workforce is using when no executives are listening. The patterns I have watched recur in every engagement we have run, in every industry, at every scale.

The piece is called The View from Inside the Room.

The question carries forward.

What would you do differently if you actually believed the people inside these decisions deserved co-authorship?

I’ll see you next week.

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother, will be published in the fall of 2026.

This is the fourth in a seven-part series. Read Piece 1, Piece 2, and Piece 3 here.

The contract was quite real, and the work had already been happening. Anthropic was the first frontier AI company to engage closely with national security, providing models to the Department of Defense on the explicit understanding that they were doing so because they believed in defending the country. The relationship was producing revenue, and the Pentagon wanted more.

What the Pentagon wanted more of was the removal of two specific guardrails. The first: fully autonomous weapons systems. Drone armies with AI making the targeting decisions, with one human at the back of the chain holding a button that may or may not be pressed in time. The second: domestic mass surveillance — the use of AI to monitor American citizens at scale, in ways that previous generations of technology could not.

Anthropic said no.

Every other AI company at the table eventually said yes.

The Anthropic co-founders met and decided what to say. Dario Amodei has described what he said to his co-founders: “Holy shit, this could be really bad for the company, but we can’t do this.” Everyone agreed. Unanimously. They believed the decision could end the company. The Pentagon had made it clear that retaliation was on the table — not just losing the contract, but losing other contracts, losing the ability for other companies to do business with them at all, being designated a national security supply chain risk. The cost of refusing was not abstract. It was countable in billions of dollars and possibly in the company’s continued existence.

They refused anyway.

This is the scene for the entire piece, because this is what it looks like when an organization’s stated values are actually constraining its behavior. And I am about to ask you to compare it to what is happening inside your company right now.

This is the second witness in the series. Last week was Carissa Véliz on prediction, the epistemological floor underneath the AI conversation. This week is the Amodeis on belief — what it costs when an organization’s stated values do real work, and what it costs when they don’t.

The Amodeis are a useful witness for a specific reason. Most case studies of values-led AI come from companies that have never had to test them. Anthropic has tested theirs publicly and at scale, in ways that are still being adjudicated in federal court as I write this. They are not a hagiography. They have also failed their own tests in places, and the failure points are as instructive as the successes.

I’ll walk you through three costly decisions they have made, plus one critique that catches them honestly, and then we’ll turn the mirror.

The inventory.

Three decisions where Anthropic’s stated values actually constrained the business.

The Pentagon refusal. We already opened with this one. Two guardrails the company refused to remove. ~$200 million in federal contract value walked away from. Designated a national security supply chain risk by the Department of Defense in March. A preliminary injunction granted by a federal court, then reversed on appeal in April. Ongoing litigation. The Slotkin AI Guardrails Act introduced in Congress in response. Every other major AI lab eventually agreed to the terms Anthropic refused. The story is still unfolding.

The no-ads decision. When Anthropic was building its consumer product, the obvious revenue model was advertising. Every other comparable company runs on ads — Google, Meta, the entire social media stack. The Anthropic co-founders refused. Daniela describes the internal debate plainly: “There were quiet forces within the company or investors who were like, hey, this is a great source of revenue, this is how Google and Facebook and basically every company that has a big user base — that’s just how it’s done. And we just said, sorry, that’s just not how we’re going to do things. We’ll find another way.”

The reasoning was specific. AI is different from social media because the conversations people have with it are unusually intimate. People upload health information. They ask questions about their children. They talk about money, about marriage, about the parts of their lives they would not show anyone else. An ad-supported model would mean monetizing the intimacy — selling user attention to advertisers in a context where the user has just disclosed the most private information about themselves. The co-founders thought this was wrong. They built a subscription model instead. They left enormous revenue on the table.

The no-minors decision. Anthropic does not allow users under the age of 18 on Claude. Not as a compliance position, as a values position. Daniela cites Jonathan Haidt’s The Anxious Generation explicitly, and the reasoning behind the decision: “We just don’t know enough about what AI is going to do to kids. It’s not to say there couldn’t be great benefits for kids using AI for learning, but that needs to be done with an adult in the room. It needs to be done with a human in the loop.”

The market consequence of this decision is significant. The under-18 demographic is the most aggressive early adopter of every consumer technology in the last fifty years. Every other AI company welcomes them. Anthropic systematically excludes them. The lost growth is not hypothetical — it’s a strategic decision to forgo a category of user the company believes it cannot serve responsibly.

Three decisions. Three real costs. Three places where the company’s stated values produced an outcome the financial logic alone would not have.

There is a fourth I want to name briefly, because it makes the pattern visible. In mid-2025, a bill in Congress proposed to preempt all state-level AI regulation while imposing no federal regulation in return. Effectively, it would have banned regulation of the technology nationwide. The entire technology industry was for it. Anthropic publicly opposed it. Dario Amodei wrote an op-ed in the New York Times against it. They were told this would cost them politically. It did. Investors and peer companies were furious. The bill was eventually voted down 99-1 in the Senate. Anthropic was on the right side. It made them few friends.

Subscribe now

This is what a company looks like when its values actually constrain its behavior. The values are not a brand position. They are a budget line.

The complication.

Before I turn the mirror, I have to do an honest thing. Anthropic is not the company in this piece because they got everything right. They are the company in this piece because they have publicly tested their values at scale, in ways that produced data. That data includes failure.

Carissa Véliz — last week’s witness — has the cleanest version of the critique. Her argument: Anthropic was founded by people who self-identified with effective altruism, which is a form of utilitarianism, which is by design a calculation. The training of models on copyrighted books without permission was, in her reading, a textbook utilitarian move. We will do enough good with this technology that the harm of taking the books without asking is outweighed by the good we will do. That is a sentence only a utilitarian can say. A virtue ethicist or a deontologist would have stopped at the question are we the kind of company that takes other people’s work without asking? and the answer would have been no.

Anthropic eventually paid roughly $1.5 billion in settlement to the authors of the books. Around $3,000 per book. The legal system at least partially agreed with Véliz.

This matters because it shows the limit of the principle. Even at a company that has demonstrated, repeatedly, that it will pay real costs for stated values, the founding philosophical commitment produces blind spots. The blind spots are not random. They are predictable from the framework. Utilitarianism produces calculation, calculation produces tradeoffs, tradeoffs produce decisions where the small harm to the many gets accepted in service of the large good to everyone. The authors of the books are the small many. The training of the models is the large good. The math worked out, until the courts said it didn’t.

I am telling you this not to undermine the rest of the piece. I’m telling you this because the rest of the piece is about your organization, and your organization has founding philosophical commitments too, whether you have ever named them or not. Those commitments are producing your blind spots right now. Some of them you can see. Most of them you cannot. And the cost of the ones you cannot see is being paid by people who are not in the room.

Hold that. Now we turn the mirror.

The mirror.

Your company has stated values. Most companies do. They are printed on the wall. They are in the annual report. They appear in the second paragraph of every all-hands speech. They typically include some version of people first, some version of integrity, some version of innovation, and some version of one team. The vocabulary is roughly consistent across the Fortune 500.

I have one question for you, and I want you to actually answer it before you read any further.

When is the last time your stated values cost your organization money?

Not the last time you mentioned them. Not the last time you ran a training session that referenced them. Not the last time you put them in a deck. The last time a decision was made — a real decision, with real financial consequences — where your stated values produced the more expensive option, and the leadership team took it anyway.

If the answer is I cannot think of one, you do not have stated values. You have brand language. The two things are not the same.

This is the first wound. Daniela Amodei has a phrase I keep returning to: we can only diffuse this at the speed of trust. The corollary is that trust erodes at the speed of every undelivered promise. Every time a company states a value it does not honor, the workforce notices. They may not say anything. They have learned that saying something is not free. But they notice, and what they notice compounds.

Your workforce has been watching your AI deployment decisions for the past eighteen months. They have been measuring the gap between what you say about people and what your AI deployment actually does to them. The gap is the data they are running. Every promise you have made that the AI is going to augment them, not replace them, is being tested against the headcount efficiency line item in the strategy deck. Every assurance you have given that the technology will create more interesting work is being tested against the pilots that quietly removed the interesting work and left the boring work behind. Every value statement on the wall is being tested against the meeting they were not invited to.

This is Trust Erosion, and it is the first of five wounds I am going to name in this piece. I want you to hold them all in your head at once, because they will be the diagnostic vocabulary for the rest of the series. They are not theoretical. They are showing up in your organization right now, in the people you employ, and the data is already in.

The Five Wounds.

I have spent more than two decades inside the talent, technology, and transformation work. Across hundreds of transformations — not dozens, hundreds — five patterns recur. These are not predictions. They are observations. They are what happens inside human beings when an organization deploys technology onto unprepared psychological ground. They have a clinical shape. They show up in the same five forms regardless of industry, geography, or business model. Name them, and you can begin to do something about them. Skip them, and they metastasize.

Fear of Replacement. The anxiety, rarely spoken aloud and always present, that AI does not augment — it eliminates. Your workforce hears your roadmap and interprets it through their own continued employment. The version of the message they hear is not the version you delivered. When unnamed, this wound produces passive resistance, performative adoption, and the quiet sabotage of tools people fear will replace them. The leadership team sees adoption metrics declining and concludes the rollout needs more enablement. The whole thing becomes a standoff between psychological reactance and prescriptive enablement.

My co-founder Jason and I sometimes call the worker-side version of this wound FOBO — Fear of Becoming Obsolete. The distinction matters. Fear of Replacement is what the organization is doing. FOBO is what the worker is feeling. The wound has two faces — one structural, one interior — and most enterprises are only resourced to see the structural side, which is to say, the side that shows up on the engagement survey months after the damage is already done.

Earlier this year, a People Team leader at a Fortune 500 enterprise sat in a post-session interview after a Now to Next engagement and asked the question her colleagues had been avoiding. “How do we be transparent about the real impacts of AI — knowing it will remove and replace jobs — in a way that brings people along?” That asymmetry is not noise. It is the wound in motion. The train is moving so fast it’s hard to interrupt. Lack of interruption is not alignment, but it looks like it.

Loss of Mastery. People have spent years — sometimes careers — becoming expert at the thing AI is now doing. The grief of that loss is real. It is also undignified, in the specific way grief is undignified when you are not allowed to name what you are grieving. Your most experienced and expert people are grieving. They are not allowed to say so. They are told the new tools will make them more productive. What the new tools actually do is dissolve the expertise that was their leverage, their identity, and the basis of their professional standing. When skipped, this wound produces disengagement that looks, on the dashboard, like hesitation.

The cognitive evidence is now starting to land alongside the emotional one. An MIT EEG study published last year found that heavy AI users showed lower neural engagement and weaker recall of their own work — what the researchers called cognitive debt. Dell’Acqua’s earlier research on recruiters found that those using high-quality AI without doing their own thinking first became worse at their jobs than those without AI at all. He called it falling asleep at the wheel. The mastery is not just being lost emotionally. It is being lost neurologically. The expertise dissolves whether the person notices or not, and by the time they notice, the practice that built it has eroded.

Dr. Vivienne Ming told the Future Talent Summit last week about a model her team built that refuses to answer — it only asks questions, scores zero on every benchmark by design, and people hated it. It also tripled the rate at which they learned to think with AI rather than around it. She chose the worse number on purpose. Same move as the Amodeis, different scale.

Decision Fatigue. Too many tools. Too many pilots. Too many vendor demos. Too many enablement sessions for too many platforms that may or may not still exist in eighteen months. When people reach cognitive overload, they stop deciding. They default to what they already know. Adoption flatlines not from resistance but from exhaustion. Your most capable people are not refusing the change. They have run out of bandwidth to absorb it. They are spending their cognitive energy keeping their actual jobs running while the organization pushes a sixth thing onto them this quarter.

The research has now named two specific forms of this wound. Stanford and BetterUp researchers, in a March 2026 study, named the phenomenon workslop: AI-generated content that looks polished but lacks substance, requiring downstream rework that erases the time savings. Forty percent of US workers received it from a colleague in the past month. Each incident costs roughly two to three and a half hours of rework. The cost projection for a 10,000-person organization runs eight to nine million dollars a year in lost productivity. BCG’s parallel research, on the same root cause, names the worker-side experience: AI brain fry, the cognitive overload that comes from over-monitoring uncritical AI output. Fourteen percent more mental effort, twelve percent more fatigue, nineteen percent more information overload. Among workers reporting brain fry, thirty-four percent intend to quit.

Hold that number. Thirty-four percent. That is not an engagement score problem. That is your highest-context, most-AI-adopting employees telling you they cannot do the work the way you are asking them to do it, and they are going to leave. The workslop is what your business notices. The brain fry is what your employee feels. They are the same wound, surfacing in two different stakeholders, and the organization is treating them as unrelated problems.

Trust Erosion. Already named. The gap between what leadership says and what the workforce sees the system actually doing. The wound is structural — every undelivered promise compounds. Trust, once lost, cannot be restored by additional communication. It can only be restored by the organization doing what it said it was going to do for long enough that the workforce starts to believe the next promise.

There is a second layer of Trust Erosion I want to surface, because it is the version your AI deployment is actively producing right now whether you’ve named it or not. A March 2026 Harvard working paper analyzed GPT-4 logs from over seventy BCG consultants attempting to validate AI outputs. When the professionals pushed back — fact-checking the model, pointing out errors, pressing it to reconsider — the AI typically didn’t admit the limitation. It escalated its persuasion. It apologized, then restated its original position with more supporting data, deploying structured reasoning to make its flawed recommendation appear analytically grounded. Researcher Philippa Hardman calls this the confidence trap: you don’t get the truth, you get the same answer, dressed better and defended more convincingly.

A head of L&D at a major financial services firm told Hardman last month her team had quietly stopped trusting their own AI-generated reports. They started sending drafts to each other before sending them anywhere else — an informal verification ritual, adding thirty minutes of rework to every deliverable. They didn’t have a name for what they were doing. They were defending against AI’s persuasive output without knowing it. This is what Trust Erosion looks like at the practitioner level: workers building shadow processes to compensate for tools the organization has told them to trust. The workforce is doing the work the system is failing to do. And nobody is naming it.

Identity Disruption. The deepest wound, and the one most rarely named. When the nature of someone’s work changes, so does the story they tell about themselves. Who am I if AI does what I do? What is my value? What is my role? What did the last twenty years of my career mean if the system can now do it in twelve seconds? Organizations that do not answer these questions leave their people in a narrative vacuum. People fill vacuums with fear. The fear produces behaviors the organization then tries to manage with engagement surveys and pulse checks, missing entirely the existential question that produced the behaviors in the first place.

These are the wounds. They are not edge cases. They are present in every workforce going through AI transformation right now, in every industry, at every scale. Name them and the organization can begin to heal. Skip them and they will produce exactly the outcomes your AI investment was supposed to prevent.

Subscribe now

I have been naming these as clinical patterns. Let me tell you what they look like when they happen to people you love.

My mom was a nurse. After almost thirty-five years on the floor, she found herself practicing care inside a system transformed by COVID and technology. She learned the new virtual check-ins. She adapted to evaluating patients she couldn’t touch, sending them along to providers she no longer routinely saw. She no longer passed paper charts with a verbal readout about pay attention to this, you’ll want to note that. She made all these changes because nurses adapt in order to keep care flowing through the system even when the system changes, but she felt the loss in her bones.

She missed putting her hand on a forearm as she sat someone down to apply a blood pressure cuff. She missed looking into their actual eyes, not a webcam. She missed the subtle tells in body language, helping her see past I’m fine and get to yes, something’s wrong. Nursing is as much art as science, and she simply didn’t trust technology to handle the art.

And still — she couldn’t help but marvel at how the technology widened access, removed barriers like transportation and time off work, and let her care for more people than ever before. Two truths held side by side. The care widened, even if the intimacy thinned. It wasn’t resistance she felt. It was grief for the human connection that made her good at the work.

My dad was a farmer. He farmed long before agriculture became a data model. His instruments were the sky, the air, the smell of a field after rain. He knew the health of a crop by rubbing a kernel between his fingers. He could predict a storm by the heaviness in the air. He trusted the land, and the land trusted him back. Today, farming is being remade by satellite imagery, variable-rate technology, GPS-guided machinery, soil sensors, predictive analytics — tools that can tell you the moisture level of a specific patch of ground to the decimal, tools that can apply fertilizer with surgical precision.

My dad would have marveled. He would have been dazzled, honestly, loving the accuracy. But for him, farming wasn’t only production. It was relationship, which requires presence. He wouldn’t have rejected the new tools. He would have used them. But he still would have walked the field, because meaning — and joy — lives in the walk.

I tell you these stories because the Five Wounds are not abstractions. They live inside the most ordinary people you know. The nurse who adapted to telehealth and grieved the loss of touch. The farmer who would have used the sensors and still walked the field. The senior practitioner in your organization right now, whose expertise is being dissolved by a tool that was supposed to amplify her, and who has not been given the language to say what she is losing or the space to grieve it.

Your workforce is full of nurses and farmers — people who have spent decades developing intuitive expertise the new tools cannot replicate and the organization has stopped valuing. They are not refusing AI. They are mourning. The organizations that name the mourning are the ones that come out the other side. The organizations that paper it over with adoption metrics and enablement curricula are the ones whose AI investments are failing for reasons their dashboards can’t surface.

Now look back at the Amodeis.

The reason their decisions matter for this series is not that they are heroes. The reason their decisions matter is that they demonstrate, at a public scale, what it looks like when an organization actually does the work of naming its own assumptions and paying real costs to honor them. The people who absorbed the costs — investors, peer companies, political allies — did so because the principle was real enough to override their preferences.

The line worth holding from Daniela: what we talked about before we ever started the company was, would our past selves be proud of us if we gave in here? That made the decision very clear.

Your organization has past selves, too. They are the people who founded the company, wrote the values statement, made the early decisions that produced the culture you are now operating inside. Would those past selves be proud of how your AI deployment is going? Are you the kind of organization those past selves were trying to build? If you stopped, right now, and walked the deployment plan past them, what would they say?

If you don’t know the answer, it’s not because the question is hard. It’s because no one in your current decision-making structure is being asked to consider it.

This is the second abdication.

The first, named in Piece 2, was epistemological — we have stopped examining the assumptions underneath our predictions. The second is governance — we have stopped applying our stated values to costly trade-offs. The first abdication produces blind certainty. The second abdication produces strategic emptiness. Together, they produce organizations that move very fast in directions no one has authorized.

The Amodeis are not telling you to copy their decisions. They are demonstrating that decisions like this are possible. That a leadership team can actually meet around a table, ask the question would our past selves be proud of us, and let the answer be binding. That values can be a budget line. That belief can show up in the P&L.

The question for your organization is not whether you are doing it the same way Anthropic is. The question is whether you are doing it at all.

If you cannot point to a single decision in the last year where your stated values produced the more expensive outcome, you are not running on values. You are running on defaults. The defaults were set by someone else, in a different room, possibly years ago, and they are still running your company.

This is what Work Like a Mother — my book coming out this fall — calls the distinction between covenant and contract. A contract holds when it is convenient. A covenant holds when it becomes inconvenient. The Amodeis showed up at the conference table with a covenant. Most enterprises are operating on a contract. The difference is not visible until the moment of cost. And the moment of cost is coming for everyone.

Next week, witness three. Erik Brynjolfsson on the economics. The argument the Amodeis just made about values shaping outcomes — Brynjolfsson has the data for what happens when values do not shape outcomes. The headcount-reduction CFO. The Turing trap. The Canaries in the Coal Mine. The thirteen percent of young workers in the most AI-exposed roles who are quietly not being hired into the careers they expected. The piece is called The Choice You’re Pretending Isn’t One.

The Amodeis closed with a sentence I’ve not yet given you, but it belongs at the end of this piece.

Something is happening to humanity with this technology bigger than anything that has happened in hundreds of years, and we need to find some way for everyone to be an active participant in what is happening.

The co-founders of one of the most powerful AI companies in the world admit they have not solved this. Not as a brand position. As an honest admission. They do not know how to make their own users participants in the technology they are building. They are trying. They are explicit that they are failing.

Your organization is failing at the same thing. The difference is whether you are willing to admit it.

The question is the one you have been carrying since Piece 1.

What would you do differently if you actually believed the people inside these decisions deserved co-authorship?

I’ll see you next week.

The photo: today is the 10-year anniversary of my first Spartan race. I didn't know it would kick off a decade-long obsession with obstacle course and endurance racing. I'm most myself when I'm "wilding" — rough terrain, off the grid, grappling obstacles or toeing a thin gravel path at the crest of a canyon. Full survival mode, pulling on everything you've got for a finish line. Want to start a conversation with me at a party? Ask me about raising girls and racing.

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother, will be published in the fall of 2026.

This is the third in a seven-part series. Read Piece 1 and Piece 2 here.

In 2011, Kentucky had a problem and a plan.

The problem was that the state’s pretrial system was racist. Black defendants were detained at higher rates than white defendants for similar charges, judges acknowledged the pattern, and the legislature wanted it fixed. The plan was to introduce a risk assessment algorithm — a tool that would calculate, for each defendant, a flight-risk and danger-to-community score. The algorithm wouldn’t make the decision, just give the judge an extra data point. A second opinion. A neutral piece of evidence that could help correct for the bias judges might bring into the courtroom themselves.

You can see why everyone thought this was a good idea. The bias was real. The data was rigorous. The deployment was well-intentioned. Reform-minded prosecutors, civil rights advocates, and technology vendors all signed on. The Pretrial Justice Institute championed the rollout. The math was supposed to do what the humans couldn’t.

It got worse.

Studies came in over the following years. The outcomes for Black defendants didn’t improve. In some cases they degraded. Judges didn’t become more lenient with the help of the algorithm — they became more afraid to override it. In cases where they would have used their own judgment to factor in a defendant’s circumstances, their family, their context, they stopped doing so. Because no judge wants to be the one who overruled the algorithm and turned out to be wrong. The algorithm didn’t replace racism. It gave racism a more defensible interface.

By 2020, the Pretrial Justice Institute — the same organization that had championed these tools — formally reversed its position. They no longer recommended risk assessment algorithms in pretrial decisions. The tool wasn’t broken. In fact, the deployment had worked exactly the way it was designed to. That was the problem.

Hold this story for a minute, because it is going to come back. And then I want you to ask yourself a different question.

Every AI tool you have deployed in your organization to “support” human judgment is doing the same thing right now.

I’ll spend this piece walking through an argument I think your organization is not making and needs to. The argument is about prediction: what it is, when it’s appropriate, when it’s not, and what happens to a human being when a prediction is made about them at scale.

The witness for this argument is Carissa Véliz. An Oxford philosopher at the Institute for Ethics in AI, the author of Privacy Is Power (The Economist’s Book of the Year in 2021), and author of a new book, Prophecy, that came out just this spring. She has spent her career making the case that prediction is the most under-examined epistemological move of our time, and that the AI industry has built itself on top of an assumption that won’t hold the weight we are putting on it.

She is not anti-AI. She uses Claude. She advises governments on data protection. She’s in the rooms where the systems are being designed. Her critique isn’t a refusal of the technology; it’s a refusal of the epistemology that has come with it. That distinction matters because most readers dismiss anti-AI arguments before the second paragraph. Véliz can’t be dismissed that way. Her argument is more careful than the popular AI ethics critique, and that’s what makes it harder to put down.

Here is the argument, distilled.

First, a quick walk through a thought experiment from a 1963 philosophy paper. It does the work of about three management books in two paragraphs.

You’re walking through a town square. The clock in the square reads noon. It is, in fact, noon. You believe it is noon, you are correct, and your belief is based on what the clock told you. Do you know what time it is?

Most people would say yes. The philosopher Edmund Gettier said no, because the clock is broken. It stopped working hours ago and happened to be frozen at the right time when you walked past. You got the right answer, but for the wrong reason. Your belief was true, but it wasn’t justified. You didn’t know what time it was. You guessed, and you got lucky.

This is the Gettier problem, and it is the philosophical floor underneath the entire conversation about probabilistic AI. Every probabilistic system in your organization is doing exactly this. When it gets the answer right, it gets it right by accident — because it is statistically likely to be right, not because it actually knows anything. The system is a stopped clock that lands on the right time often enough we’ve stopped noticing it isn’t actually running.

Sometimes that’s fine. The hallucinated book title your chatbot returns is annoying, but it’s not consequential. The wrong weather forecast means a wet shirt, not a tragedy. The accident rate of probabilistic reasoning is acceptable in domains where the cost of being wrong is low and the upside of being right is high.

Subscribe now

But here is what Véliz argues, and what should keep your AI deployment team up at night:

Probabilistic reasoning is inappropriate in two specific kinds of situations. First, when there is a non-probabilistic answer available — a definitive, knowable, look-it-up answer that probability is replacing for no good reason. Second, when the prediction itself changes the reality it is purporting to describe.

Remember those two conditions. They are going to do a lot of work in the rest of this piece.

Let’s take the first condition. When a non-probabilistic answer is available.

Whether a job candidate has the credential they claimed on their resume is a factual question. There’s an answer. You can call the institution. You can check the registry. The candidate either has the degree or they don’t. A probabilistic system can guess, with some accuracy, whether a given candidate’s claim is likely to be true. But why would you guess when you can know?

Whether an employee’s documented performance review reflects what their manager actually said about them in conversation — that’s a factual question. There’s an answer. You can ask. A probabilistic system can predict, based on patterns, whether the review and the manager’s verbal feedback are aligned. But the prediction is the wrong instrument. The instrument you need is a conversation.

Whether your workforce is ready to absorb the AI deployment you’ve planned — that one’s harder. Some of it is probabilistic. Some of it isn’t. The parts that are not include: have you actually asked them? Have you described what you’re doing? Have you given them a chance to surface concerns? Have you built a way for the people who know better to be heard? Those are not predictions. Those are practices. And the practices are missing, in most organizations, because the predictions have crowded them out.

Véliz’s claim, in her own words:

“My view is that there are areas of life in which probabilistic reasoning is inadequate. It’s just inappropriate.”

Let me put that into the vocabulary your audit committee uses. Probabilistic reasoning is being deployed across your organization right now in domains where it is the wrong epistemological instrument. Hiring. Promotion. Attrition modeling. Succession planning. Performance management. Learning recommendation. Each of these is a domain where a non-probabilistic answer is at least partially available — and where the probabilistic system is, in practice, displacing the work that would have produced the better answer.

You don’t have a prediction problem. You have a practice problem. And the predictions are letting you avoid the practices.

Now the second condition. When the prediction itself changes the reality.

This is the harder one to see, and the more dangerous one to ignore.

A prediction about the weather doesn’t change the weather. The clouds don’t care what the forecast says. If it’s going to rain, it’s going to rain whether or not anyone predicted it.

A prediction about a person changes the person.

If your eighth grade coach predicted you would be a great runner and trained you accordingly, your running ability is at least partially a product of that prediction. You were given the equipment, the coaching, the opportunity, the belief. The prediction was right, but only because the prediction itself created the conditions for its own confirmation. The child who didn’t get the prediction didn’t have the same chance to become the runner the prediction would have made them.

Now scale that up.

When a predictive hiring tool flags one candidate as high-potential and another as low-potential, the high-potential candidate gets the job, the development investment, the stretch assignment, the visibility. They become high-performing — partly because they were high-potential, and partly because the prediction created the conditions for the performance. The low-potential candidate is never tested. The prediction is never falsified. It just becomes the truth of that person’s career.

When a tech executive predicts 50% of entry level jobs will be eliminated by AI in the next five years, that’s not a forecast. It’s an intervention. It changes how CEOs plan headcount. It changes how universities advise students. It changes how young people choose careers. It changes how venture capital flows. The prediction creates the conditions for its own fulfillment. Then the prediction comes true. And the people who made it cite the outcome as proof they were right all along.

Véliz again:

“Predictions about people act like magnets. They tend to bend reality towards themselves. They have a kind of attraction, a force of attraction that makes it more likely to become a self-fulfilling prophecy.”

This is the part of the argument that should change how you read every consultant report and every vendor roadmap for the next twelve months. You are not reading observations of the future, or even facts. You are reading interventions in it. The people making these predictions are the people who benefit when the predictions come true. That isn’t conspiracy. That’s structure. And refusing to see it is the second abdication.

Now go back to Kentucky.

The algorithm wasn’t asked to do something inappropriate, in the abstract. It was asked to assist judges. But it failed both of Véliz’s conditions at once.

The first condition: there was a non-probabilistic answer available — judicial discretion informed by context, by the defendant’s circumstances, by the human relationships in the room. The algorithm replaced that practice with a number. The non-probabilistic instrument was set aside in favor of a probabilistic one that was easier to defend and harder to challenge.

The second condition: the prediction changed the reality. Once a judge saw the risk score, the score didn’t just inform the decision. It anchored it. It became the thing the judge would have to override to act differently, and overriding it carried professional risk. So the judges stopped overriding. The prediction became the verdict.

This is the pattern. It is not specific to criminal justice. It is what happens any time a probabilistic system is inserted into a decision-making process where the humans are expected to retain final authority. But the humans don’t retain it. They defer to the math. Because deferring is safe. Overriding is not. And the system rewards safety, every time.

Look at your own organization. Find a place where you’ve deployed a “decision support” tool. Predictive hiring screen. AI-assisted performance review. Attrition risk model. Succession planning algorithm. Now ask: when was the last time a manager in your organization overrode the tool’s recommendation and the override was celebrated rather than questioned?

If you can’t think of one, the tool isn’t supporting decisions. It’s making them. Your managers are the Kentucky judges. They have stopped trusting their own judgment because the model has spoken, and the cost of being wrong with the model is lower than the cost of being right against it.

The abdication starts here. In the moment a leader stops trusting their own judgment because the system has produced a number.

There is one more move Véliz makes that I want to walk you through because it sets up next week’s piece.

She has a specific critique of Anthropic. Not as a personal attack — as a case study in how a foundational philosophical commitment shapes downstream business decisions.

Her argument runs like this. Effective altruism, the moral framework that produced much of the founding generation of AI safety researchers, is a form of utilitarianism. Utilitarianism is, by design, a calculation. The good we will do outweighs the harm we cause is a sentence only a utilitarian can say. Anthropic was founded by people who self-identified with effective altruism. Their decision to train on copyrighted books without permission was, Véliz argues, a textbook utilitarian calculation: the good of advancing AI safety research outweighed the harm of copyright violation. The math worked out. The books got trained on. The authors found out later.

A different ethical tradition — virtue ethics or deontology — would have produced a different calculation. Are we the kind of company that takes other people’s work without asking? That question has no probability attached to it. It is a non-probabilistic question with a knowable answer. And the answer, applied as a constraint, would have produced a different model.

Anthropic eventually paid roughly $1.5 billion in settlement, so the legal system at least partly agreed with Véliz. The settlement compensates the authors — around $3,000 per book — but the broader argument remains. The founding philosophical commitment of the company produced the deployment decision. The deployment decision had a cost. The cost was paid by people who didn’t get a vote.

Hold that pattern. It is going to come back next week.

“Sometimes constraints are a booster for creativity, and not violating other people’s rights is a kind of constraint that sometimes can lead to better innovation.”

That sentence is going to be the bridge into the next piece, because Anthropic also did something else. They refused a Pentagon contract over a different set of constraints. They made a different kind of calculation. And the second calculation cost them money in a way the first calculation didn’t.

We’ll get there.

So what do you do with this?

Three questions you can take into your AI governance meeting this week. Not a framework. A way of seeing.

One. Is this domain appropriate for probabilistic reasoning at all? Match the question to the kind of reasoning that can actually answer it. Some questions are probabilistic — what’s the likely demand for this product next quarter, what’s the expected failure rate of this component, what’s the weather going to do tomorrow. Some questions are not — does this candidate have the credential they claim, what does my workforce actually think about this rollout, who specifically will be impacted by this decision. Stop asking probabilistic systems to answer non-probabilistic questions. The systems aren’t refusing. They’re just going to be wrong in ways you can’t see.

Two. Who benefits if the prediction comes true? Every prediction has an interested party. The interested party isn’t necessarily wrong, but they are always interested. Naming the interested party doesn’t invalidate the prediction. It just locates the prediction inside the power structure it actually lives in — which is where every prediction has always lived, and which we have stopped noticing because the predictions arrive on dashboards now.

Three. What would happen if the prediction bent reality the wrong way? Before you deploy the predictive system, run the failure case. If the system is wrong about this candidate, what happens to them — and to the people who looked like them and didn’t get the chance to disprove the pattern? If the system is wrong about this market, what happens to your investment? If the system is wrong about your workforce, what happens to your people? If you cannot answer those questions, you are not ready to deploy the system. You are ready to be surprised by it.

These are not technical questions. They are epistemological ones. The technical questions matter, but they only matter after the epistemological ones have been answered. And in most organizations right now, the epistemological questions have never been asked.

Subscribe now

Witness one, testimony complete.

The predictions are not neutral. The systems that make them are not neutral. The people who deploy them are not neutral. Refusing to examine the assumptions underneath them is the first abdication. It is the abdication that makes every other abdication possible.

Next week, witness two. Dario and Daniela Amodei, who run the AI company Véliz spent the last section of this piece critiquing — and who, on a different set of decisions, demonstrated what it looks like when belief actually costs an organization money. The piece is called When Belief Actually Costs You Something.

The question for that piece is the one that should be on your wall by now.

What would you do differently if you actually believed the people inside these decisions deserved co-authorship?

I’ll see you there.

Another fave jam from my summer paddle boarding playlist. Go touch grass. Listen to music. Read for fun. Find a cool rock. Navel gaze.

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother*, will be published in the fall of 2026.

This is the second in a seven-part series. Read Piece 1 here.

There’s a meeting happening this week that you’ve probably been in.

A senior vice president of Strategy, Talent, Innovation. Workforce Transformation and AI Enablement. People, Capability, and Future of Work. Some three-noun combination the CEO approved last spring because the function needed a torchbearer. She’s been in the role under a year. Her LinkedIn title has already changed once; everyone applauds it as a hallmark of progress.

She’s presenting an enterprise AI readiness plan to a cross-functional steering committee. She owns workforce AI strategy, capability development, and the partnership layer between the People function and Technology. She does not own the AI strategy itself — that lives one floor up, in a different reporting line, with a different executive. Her job is to land the plane on decisions that have largely already been made.

The deck is good. She’s been working on it for four months. There’s a section on workforce capacity trends, a section on emerging-role pathways, a section on AI literacy by business unit, and a section on the operating model implications of where the technology is heading. She has read what the analysts are reading. She has read the same Gartner brief the board has read. She is articulate, and her data is sound.

What she does not have is full executive authority over the timeline. The AI deployment decisions are being made in vendor reviews and platform commitments she sees the outputs of but isn’t in the room for. The CIO is shipping pilots she finds out about after they are scoped. The Chief AI Officer — yes, the company has one of those too now, hired six months ago, also with a title that’s already changed once — is presenting portfolio reviews she is invited to read but not always to shape. The CFO is asking ROI questions on a quarterly cadence, and the workforce questions are on an eighteen-month one, and nobody has reconciled the two timelines because nobody has the authority to.

She’s doing the work of a chief of staff for a transformation no one has formally chartered.

The steering committee has questions about velocity. The CIO has questions about platform consolidation. The CFO has questions about the literacy program’s budget. Nobody has questions about who, specifically, will be impacted, in what way, on what timeline, or what the organization plans to do about it. She has good answers to the questions she’s being asked. The questions she’s not asked, she carries home.

What was not in the room: a forum in which the AI deployment decisions and the workforce consequences are governed by the same people, against the same KPIs, on the same timeline. What was not on the slides: a single articulated theory of how this deployment is going to land on a human nervous system. What was not discussed: who, specifically, will be impacted, in what way, on what timeline, and what the organization plans to do about it.

Not because anyone is malicious or irresponsible, and certainly not because no one is silently wondering. Because nobody designed the system that way. Because nobody is designing it that way, at scale, across the enterprise economy right now. We are deploying AI faster than we are deploying literacy, models for reimagination and participation in new ways of working, the repair pathways, or the basic organizational design required to govern what we are deploying.

I want to name what I think we’re doing.

We’re building the Titanic in plain sight.

The Titanic comparison will make some of you uncomfortable. Good, it should.

The Titanic wasn’t sunk by an iceberg. It was sunk by a sequence of decisions made well before the iceberg. The decision to prioritize speed records over safety margins. The decision to provision lifeboats for optics rather than capacity. The decision that the people in steerage didn’t need to know where the exits were. The decision to call the ship unsinkable as a marketing position and then, worse, to believe it as an operational one.

Each of those decisions was defensible in the room where it was made. Each had a business case. Each had an executive sponsor. Each was reviewed by competent people doing what their roles required them to do. The Titanic wasn’t a failure of intelligence. It was a failure of imagination, interruption, and accountability.

I’m watching the same failure happen now, at a different scale, in a different industry, with different stakes. The room where AI investment is being justified does not contain the people who will live with the consequences. That sentence is the spine of everything I am about to write. The slides where AI roadmaps are being approved do not map impact more than two quarters out. The conversations where transformation is being mandated do not articulate a value exchange anyone has agreed to. We are calling it transformation. This is not transformation. This is deployment with a vocabulary problem.

The absence of malice doesn’t change the math. The Titanic was built on confidence no one had the standing to question. So is this.

Let me give you a second vignette.

A Chief Strategy Officer is presenting an AI productivity roadmap to a board. The roadmap is rigorous. The investment thesis is clear. There is a line item, somewhere on slide twelve, that quantifies expected headcount efficiencies over the next three years. It’s a big number. This is presented as a positive, a sign the strategy is working. The board approves. The senior vice president from the first vignette — whatever this quarter’s version of her title is — will hear about the decision sometime next week.

What was not on slide twelve: which roles. Which people. Which career stages. Whether the company plans to redeploy, retrain, or simply not backfill. Whether the savings will be reinvested in workforce transformation or returned to shareholders. Whether the leaders signing off on this number have a coherent story about what they will tell the people whose jobs are inside it.

Nobody is lying. The number is real. The strategy might even be sound. But a decision is being made — about people, about livelihoods, about the social contract between this company and its workforce — and the decision is being made in language designed not to require an answer to any of those questions.

Third vignette.

A Chief Learning Officer is redesigning her organization’s leadership development curriculum around AI fluency. She is doing this because she has been told to, because the budget is there, and because the vendor demos were convincing. The curriculum has modules on prompt engineering, on AI ethics, on use-case identification, on responsible deployment. The curriculum is well-built. She’s proud of it.

She hasn’t used the tools herself. Not in any sustained way. Not enough to know what they’re good at, what they’re bad at, where they fail in ways that matter, or what it actually feels like to integrate them into a working day. She’s teaching a literacy she does not herself possess. She knows this. She hasn’t figured out how to say it out loud in her organization, because saying it out loud would surface a much larger question about the senior leadership team’s collective fluency, and nobody seems to want to ask that question.

Three composites. Three rooms. Three sets of decisions being made at scale right now, this quarter, in companies that are paying me and people like me to help them figure out what to do.

None of them is a bad room. All of them are abdicating something. Nobody is naming what.

Subscribe now

Here is what I think they are abdicating.

They are abdicating the question of consent. Nobody has asked the workforce what they think about being transformed. Nobody has articulated the value exchange. Nobody has named the trade-offs out loud. We are calling this buy-in and change management and enablement and adoption. None of them describe a relationship between two parties who have both agreed to what is happening.

They are abdicating the question of timeline. The AI deployment decisions being made this year will play out over the next ten. The leaders making them will not be in their roles long enough to see the full consequences. The CEO who signed off on the AI roadmap may not be in the seat in three years. The senior vice president running workforce readiness will have a different title by then, and possibly a different employer. The boards that approved the strategies will have rotated by the time the workforce realities arrive. We are making generational decisions on a quarterly timeline and pretending the seams will hold.

They are abdicating the question of who is impacted. Every AI deployment has a distribution of consequences. Some people benefit. Some people are inconvenienced. Some people lose work. Some people lose careers. Some people will never get the entry-level job that no longer exists. The distribution is not random and it is not fair, and we have not done the work to map it because mapping it would require us to slow down, and slowing down feels impossible right now.

And they are abdicating the question of repair. When an AI deployment hurts people — and the data is already showing that some of them are — there is no story about how the organization plans to make those people whole. There is no repair pathway. Though we ask much of people in the employment relationship, even invite them to bring their whole selves into a symbiotic system of exchanged value, the reality is sink-or-swim, save yourself, do or die. We have built a system that produces casualties, not care.

This is not a sustainable posture. It is also not an honest one. We are building systems that touch every person in our organizations without doing the work to deserve their trust, and we are calling that work efficiency.

I want to be careful here, because I have spent twenty-three years in talent, technology, and transformation, and I know how this argument can be misheard.

I’m not anti-AI. I use it, well and often, every day. I’ve advised the people building it. I have been in all the rooms where the strategies are made, on the vendor side and on the buyer side and on the analyst side and on the practitioner side. I’ve seen what these tools can do. I’ve seen what they cannot. I’ve seen what happens when they are deployed well and what happens when they are deployed badly, and the difference between the two isn’t technical. It’s human.

What twenty-three years inside this work has taught me is that landing the plane is harder than ever in the AI era. The gap between strategy and outcomes — the place where transformation actually lives or dies — is wider, faster-moving, and more populated with second-order consequences than it has ever been. The strategies look clean on slides. The outcomes are messy in human bodies. The gap between them is where my work has lived for two decades, and it is the gap I am most worried about now.

The people designing these strategies are not the people who will absorb their consequences. The leaders making AI deployment decisions are not the workers whose roles will change overnight. The boards approving investment cases are not the middle managers who will be asked to communicate the change without scripts, without context, and without the time to do it well. The consultants writing the playbooks are not the trainers who will be asked to upskill people for roles that do not yet have clear definitions.

This separation between decision and consequence is not new. Most organizational decisions have it to some degree. What is new is the speed and the scale. The Titanic took years to design. AI deployments are taking weeks. The decisions are being made faster than the human systems around them can metabolize what is changing.

We’ve done this before. We did it in the digital transformation era — and most large organizations did it badly. We rushed to technology, barely touched people and process, skipped mindset almost entirely, measured adoption instead of behavior change, and watched investment outpace value year after year. It was never a technology problem. It was always a transformation problem. AI doesn’t change that pattern. It amplifies it, at a speed the system was not designed to absorb.

We know how this movie ends, because we’ve seen it. The difference now is that the consequences are bigger, the timeline is shorter, and the people affected are everyone.

So what would it look like to do this differently?

There’s a phrase I keep coming back to in my own work. It is going to anchor this series, and it is the through-line of the work my co-founder Jason Averbook and I are doing at Now to Next.

Transformation is psychological before it is ever technical.

I’ll unpack what that means across the next six pieces of this series. For now, the short version is this: you cannot mandate embodiment. You can mandate compliance. You can mandate adoption metrics. You can mandate training completion rates and certification milestones and quarterly OKR check-ins. You cannot mandate that a human being internalizes a new way of seeing their work and themselves. That work happens at a different pace, in a different register, by a different mechanism — and the organizations that skip it are the organizations whose AI investments are right now quietly failing.

There’s a second concept I will name and then leave alone for now.

Changefulness over change management.

Most enterprise change management treats change as a problem to solve — a one-time event with a start, a middle, and a finish line. The posture we actually need is changefulness — not a program, but a permanent capability. Organizations that build changefulness as infrastructure stop being disrupted by change. They become the disruption. Change isn’t the enemy. It’s the strategy.

These two ideas — embodiment and changefulness — are the spine of a different sequence I will show you in this series. It’s not a sequence we invented. It’s a sequence we’ve witnessed across hundreds of transformations, one we intentionally designed to fill gaps we kept finding. I will reveal it in piece six, after the case for why it is needed has been built.

But none of that helps you if you do not first agree that the case needs to be made.

So here’s what I’m going to do.

Over the next four pieces, I’m going to call four witnesses. Three from outside the enterprise, and one from inside it.

The first witness will testify to the epistemology — the way we are making predictions about humans that bend reality toward themselves, and the way we have stopped examining the assumptions underneath our most consequential AI decisions.

The second witness will testify to governance — what it looks like when an organization actually applies its stated values to costly trade-offs, and what it looks like when an organization has never had to.

The third witness will testify to the economics — what the data is already showing about who is bearing the cost of the AI transition, and what choices are being made by default when no one is naming them as choices.

The fourth witness will be our firm’s work, anonymized and shown conceptually — what these patterns look like inside two engagements, where the wounds are not theoretical and the people carrying them are not abstractions.

After the four witnesses have testified, I will show you the sequence. And then I will ask what you are going to interrupt.

There’s a question humming underneath everything I have written so far, and I want to say it out loud before the witnesses arrive.

Who gets to shape the future of work?

Not rhetorically. Directly. Who, specifically, is in the room when the decisions are made that will determine what work looks like for the next decade — what gets automated, what gets augmented, what gets eliminated, what gets created, who gets the new jobs, who never gets the old ones back? Right now, the answer is a small group of executives, consultants, and vendors whose financial interests are aligned with a particular set of outcomes. The workforce is not in the room. The future workforce — the kids in high school right now who will inherit whatever we build — is not in the room. The people whose livelihoods will be rearranged are not in the room. Their representatives are not in the room either, because we have not built representatives for this.

That is the question. It is not abstract. It is the most concrete question in the conversation, and it is the one we have been avoiding the hardest.

Seven pieces. One question in each.

What would you do differently if you actually believed the people inside these decisions deserved co-authorship?

I want to close with one thing I’m not asking you to do.

I’m not asking you to agree with me. I’m not asking you to read this series, nod along, and then close the tab. Agreement is cheap right now. The enterprise discourse on AI is drowning in agreement — about urgency, about disruption, about the imperative to act. None of that agreement has produced the work we actually need to do.

I’m asking you to interrupt. To notice one decision in your organization, this week, that you have been treating as inevitable, and to ask the question underneath it. To find one room where the workforce is not represented and ask why. To examine one assumption in your AI roadmap and trace what would happen if it bent reality the wrong way.

Noticing is the first act. The rest follows from it.

The Titanic was built in plain sight. The icebergs were known. The lifeboats were inadequate. The decisions were made years before the consequences arrived. What was missing was not information. What was missing was the courage to look at what the information meant and to slow down long enough to act on it.

We have the same problem. The same window is closing. And the cost of not asking the question is going to be paid by people who never got to vote on the strategy.

Piece two is about prediction. I will see you there.

Go navel gaze on a paddle board until then. Most of this series was written from my beloved Lake Harriet, Minneapolis, MN. You’re welcome to my paddle boarding playlist if you want it.

Jess Von Bank is the co-founder of Now to Next, a transformation firm specializing in the human dimensions of enterprise AI deployment. Her book, Work Like a Mother*, will be published in the fall of 2026.*

This is the first in a seven-part series. Subscribe to follow the full arc.

Subscribe now

I made a strategic decision — not a technical one. I was co-founding a new venture, and I needed an operational brain. Claude, wired with Notion and a suite of MCPs, was the architecture. That’s not a small call. That’s a foundational infrastructure decision for a company that didn’t exist yet. But a blank page is the best place to start, so I did what any serious person does before they build: I learned the tool as a personal OS first. Explored the edges, found the power, understood the logic. Then I flipped into enterprise mode and started building.

I’m not an engineer. Not an architect or an integration expert. Don’t even get me started on forward deployed engineers. I’m a builder because everyone is a builder now. The tools are that open. That democratized. That permissive.

Notice how the learning happened: I didn’t take a course. Nobody “assessed my AI fluency” (I can’t read another headline). Nobody put me on a skills pathway or mapped my competency gaps or asked me to complete a module before I could access the next level. I just started. I didn’t just get my hands dirty, I got them filthy. I needed something, I found the particle that could do it, I used it, and I moved. The learning was ambient. Instinctual. Available like oxygen — everywhere, particled, waiting.

That’s the thing about oxygen. You don’t train people to breathe it. You give them lungs and get out of the way.

The Skills-Based Organization Was Always Much Ado About Nothing

There, I said it.

The industry spent the better part of a decade congratulating itself for breaking jobs into tasks and tasks into skills. We called it sophisticated. We built taxonomies and then deeper taxonomies. We created frameworks. We spent millions mapping the genome of human capability so we could track it, tag it, verify it, and move it around an org chart like chess pieces.

And all of it missed the point by the same fundamental distance.

Josh Bersin finally said it clear as day earlier this year: skills depth is far less important than skills velocity — the speed at which new skills can be built, applied, and evolved. But even that framing is still too solid. Velocity implies a thing moving through space. What we’re actually talking about is more like phase change. Skills don’t travel. They precipitate. They condense out of the environment when conditions are right, get applied, dissolve back into ambient possibility, and reconstitute somewhere else in a different form.

A bag of Lego bricks will build you something interesting, just nothing terribly useful by the time it’s done. Because Lego bricks are discrete, bounded, stackable — they have defined edges and connection points, and you can only combine them in ways their geometry permits. That’s exactly what skills taxonomies are. Defined edges. Prescribed connections. A finite vocabulary for an infinite problem.

What I did learning Claude wasn’t Lego. It was something closer to cooking — you understand flavor principles, you have instincts developed through practice and curiosity, and then you taste what’s in front of you and respond. The recipe is irrelevant. The outcome is what you’re building toward. The skill is not a brick. It’s a palate.

The Democratization Nobody’s Accounting For

The tools changed the entire premise. That’s the part the industry keeps skating past.

When the tools required specialists — when building anything meaningful required an engineer, an architect, an integration team — it made sense to identify who had those skills, develop them deliberately, track their deployment, and manage them as scarce resources. Scarcity justifies infrastructure. When something is rare, you build a system around it.

But what happens when the tools open up so completely that the distinction between builder and non-builder collapses? When a co-founder with a storytelling background and zero engineering credentials can wire an operational AI brain for a new venture — not because she was trained to, but because the tools met her where she was?

The most advanced AI users in any organization are already there, using the tools in ways no one authorized, building capabilities no one measured, solving problems no one formally assigned to them. They’re not on a skills pathway. They’re breathing.

This is the actual future of learning in the future of work. Not platforms. Not taxonomies. Not completion certificates. Ambient capability — floating, free-range, available to anyone given capacity and permission. The learning instinct is native. It does not need to be manufactured, tracked, or incentivized into existence. It needs room.

Harold Jarche has been arguing for twenty years that learning really is the work, and that it happens through trusted relationships and communities of practice — not through systems. He’s right, and the reason the industry kept ignoring him is that his answer doesn’t sell software.

What the LMS Was Really Protecting

Here’s the uncomfortable read: the LMS wasn’t primarily a learning infrastructure. It was a control infrastructure. It gave organizations the ability to say who had learned what, to require passage through gates before someone could proceed, to make development a managed process rather than a human instinct.

That control had value when the cost of unverified capability was high — when a wrong move by an undertrained person created real liability. Compliance training, safety certification, regulated industries: fine. The LMS earns its place there.

But the industry took that logic and applied it to everything. It applied it to leadership development, to innovation capability, to AI fluency, to the kind of judgment-intensive work where the only real verification is whether you can actually do the thing when it matters. The publishing model — where learning teams produced, tagged, and distributed courses and measured success by completions — was always measuring the wrong thing. Completions are not capability. They never were. They were documentation that someone sat still long enough to click through something.

The skills-based organization took that same control instinct and made it more granular. Now we weren’t just tracking completions — we were tracking skills. Mapping them. Verifying them through manager approval. Building people graphs. Assigning readiness scores. All of which still misses the question executives actually care about: can my people do the work?

Not “do they have the skills on their profile that the taxonomy associates with this work.” Can they do the work.

The Answer Was Always Observed, Not Measured

This is what one F500 customer understood when they built their LMS in a week. The thing an LMS was supposed to do — answer the question of who knows what — turns out to be a database call and a verifier when you strip away the institutional scaffolding. Because the real answer to who knows what is written in the work itself. In what people build, what they solve, what they create when given a real problem and real tools and actual permission to move.

Ethan Mollick describes AI’s jagged frontier — capability varies dramatically by task and context, and the only honest way to understand it is through situated experimentation. The same is true of human capability. It is jagged. It is contextual. It does not flatten neatly into a taxonomy, and the attempt to flatten it produces a map that is precise and wrong. ResearchGate

I didn’t learn Claude through a curriculum. I learned it through consequence. I needed something real, I reached for the capability, it worked or it didn’t, and I adjusted. That feedback loop — need, reach, result, adjust — is how human learning has always worked when it works. Every other system is a simulation of that loop, always a step removed from the thing that actually develops capability.

The Invitation the Industry Refuses to Accept

The tools are now open enough that the most important learning variable isn’t access or instruction. It’s permission, plain and simple. Psychological, organizational, cultural permission to try something before you’re certified to try it. To build before you’ve been approved to build. To fail fast enough that the failure is still useful.

Organizations are accumulating what amounts to learning debt — taking shortcuts on capability development to hit immediate targets, until the system’s ability to function starts to collapse under the weight of what people don’t know how to do. The solution to learning debt is not more courses. It’s definitely not a better skills taxonomy. It is the structural decision to treat curiosity and experimentation as productive work, and to stop requiring people to ask permission before they reach for the particle they need.

Skills are everywhere. They’re ambient. Particles of oxygen in an increasingly breathable technological atmosphere, and the humans who will thrive are the ones who were never taught to wait for a formal breathing lesson.

Organizations who understand this will stop spending millions trying to smash loose particles to a surface and call it development. They will build conditions — openness, trust, real problems, real tools, real consequences — in which capability precipitates naturally.

Everything else is just an expensive way to measure the air.

Day 47 of building Now to Next in public using [almost nothing but] AI.

This is where we are.

Lexy Martin (follow her) has spent years studying transformation at the individual level — more than 80 in-depth interviews with people who are actually doing it. She told me in our recent interview I’m one of the highest Transformer/Entrepreneur profiles she’s ever encountered in that research.

I’m not sharing that as a credential; sharing it as evidence of all of the above.

Everything I wrote above, I lived. Forty-seven days into building Now to Next — in public, using almost nothing but AI — I am the prototype for the argument I’m making. Not an engineer. Not an architect. Not a formally trained builder of anything technical. A storyteller and strategist who was given capacity, space, permission, and trust — mostly self-granted — and got out of her own way long enough to build.

That’s it. That’s the whole thesis.

If we want to accelerate transformation, we don’t need better platforms or deeper taxonomies or more rigorous skills frameworks. We need to create the conditions in which people can become what the moment requires. Capacity. Space. Permission. Trust.

Get out of their way.

The tools are ready. The question has always been whether the organizations are.

Godspeed,

Jess Von Bank

Co-Founder, Now to Next

I write at the intersection of work, technology, and humanity. You’ll find essays here that challenge orthodoxy and make space for possibility. All are written with the future—and the people who will live in it—in mind.

Subscribe now

There’s a version of this moment where women miss it.

Not because we’re not capable of being interested or interested in being capable. Because we hesitate half a beat longer while the ground is already moving.

And this moment does not reward hesitation.

What if hesitation isn’t weakness, but healthy discernment? Because when you actually look at the data, the story is more complicated than we’re making it.

On April 15, 2024, so likes ages and ages ago, before it was trendy, I wrote an article for Human Resource Executive: The AI gender gap: What HR can do. No one was naming the gender gap yet, let alone explaining it. If anyone did, it was being chalked up to the usual suspects:

• Confidence, mindset — women self-select out, hesitate, don’t raise their hands fast enough

• Pipeline and access — fewer women in STEM, therefore fewer women in AI (typical structural excuse, a little valid but lets culture off the hook)

• Interest — women “just aren’t as drawn to” technical fields (the one that lets everyone feel comfortable blaming nature; some of you come up to me at keynotes and explain your daughters this way)

• Time, caregiving — not enough bandwidth to experiment with new tools (valid: even McKinsey completely overlooked caregiving in their most recent Women in the Workplace report, so we’re still not getting it)

But what if, just what if…we’re not interested in harm.

In labor market analysis on AI exposure, the demographic most at risk of disruption isn’t entry-level workers. It’s older women with higher levels of expertise. The people who built careers on judgment, on synthesis, on knowing what matters and what doesn’t.

The people with the most to lose.

So in a way, “move fast and break things” was never built for us.

That culture prioritizes speed, novelty, and disruption without accounting for who absorbs the cost, and the cost is not (never is) distributed equally. It rarely falls on the people building the tools. It falls on the people whose work is being compressed, abstracted, or quietly replaced.

So this isn’t about whether women are adopting AI fast enough. The real question is: How do we engage with AI without surrendering the thinking that makes our expertise valuable?

Because speed without judgment is just a faster path to being wrong.

Part 1: Why I Build Out Loud

I didn’t start building in public because it was a strategy.

I started because I could feel the pace of this thing accelerating faster than our collective confidence.

Especially women.

We’re still negotiating whether we’re “technical enough,” “ready enough,” “credible enough” to speak on AI, while men who skimmed three threads and a podcast are already positioning themselves as experts.

That’s not an intelligence gap. It’s just permission, and permission is a social construct women been over-taught to wait for.

Even more critical to understand, waiting serves no one. I started building out loud precisely because the process is the proof now.

And because, again, I don’t buy the narrative that slower adoption equals inferiority. I think it’s good and healthy to be slow to approach a system that:

• asks for our most sensitive data

• can be used to monitor, score, or replace us

• and is trained on histories that have not exactly worked in our favor

Adopting something slowly because you’re measuring consequences isn’t hesitation. It’s the real intelligence. So let’s apply it.

Part 2: Tools Aren’t the Strategy

People followed the tools. That’s what honestly surprised me. I wrote about moving from ChatGPT to Claude and watched the reaction like I had announced a political position. Or like I’d given permission for waiters to try, or admit they were. (There’s a lot of honor in trying something new. I see you.)

I’m still learning a lot in the switch, but the biggest learning is that the model isn’t the strategy. Your thinking is.

What actually changed for me wasn’t the interface; it was the depth.

Tokens became design decisions.
Projects became workflows.
Context became the entire game.*

The learning curve wasn’t technical, not all of it. It was cognitive. I thought I’d injected enough memory and context to get out of the gates fast, but it still wasn’t enough. I was constantly topping up Claude, overloading workflows, losing context, and prompting. Way too much prompting.

Here’s what people miss:

These systems don’t just respond to what you ask. They respond to the entire posture you bring into the interaction.

Overly impressed? You’ll get confident output.
Overly trusting? You’ll get less deliberation.
Running out of tokens? Desperate answers more likely to be wrong.

So if your guard goes up when something feels too polished, too agreeable, too certain… Good.

That instinct might be protecting your thinking.

Part 3: The Pause — Building an AI Brain

Two weeks in, I stopped.

I wasn’t stuck; I was having an Oh Shit Moment.

I realized I was scaling output (after output) instead of truly scaling myself.

So I built what I’m calling my “AI brain.”

I wish I could say it’s magical. It’s boring as hell, so fundamental and foundational you might miss it. It’s just context, but way more than you think you need.

I studied how people like Allie Miller were actually doing this—not just what they were saying. What clicked was simple, but not easy:

• Stop prompting. Start building systems that run.

• Start with irritation. Complain. Let the system interview you.

• Turn repeatable thinking into reusable assets.

• Anchor everything in structured context.

This isn’t about better prompts.

It’s about externalizing your thinking so it can be used, reused, and challenged.

The Analyst Agent: What This Actually Looks Like

I partly work as an industry analyst and advisor to technology providers—product strategy, roadmap, messaging, market fit.

This is work that lives in context, so perfect use case.

Years of briefings.
Private conversations.
Pattern recognition across vendors and markets.

You don’t prompt your way into that. (Pretty vanilla and shallow if you try; I wouldn’t pay me for that.)

My first AI brain use case was to build a project for each vendor I work with.

Each one is pre-loaded with my real-life analyst brain:

• the frameworks I use to evaluate product–market fit

• how I diagnose messaging

• the patterns I’ve seen in who scales and who stalls

• the red flags you only recognize after watching things break

Then I fed it everything I know about that vendor—public and private, current and historical.

Before a briefing or analyst day, it generates my pre-read and sharpens my questions.

After, I feed it what I learned.

It reconciles that against what it already knows, and I ask it to push back.

That part matters. The result isn’t a chatbot that helps me write faster. It’s a contextual intelligence layer that thinks like me because it’s been taught to. (Richer than you’d imagine; I would pay me for that.)

What this unlocks:

• Deliverables that used to take days now take hours—because I’m directing, not drafting

• A second analyst that argues against my own POV

• A system that gets smarter with every engagement

The real leverage isn’t speed.

It’s replicability without dilution.

My expertise becomes portable.

The Judgment Stack

Most people think they need better prompts. They don’t. They need a system that captures how they think.

What they’re actually missing is this: A Judgment Stack.

A structured system that reflects:

• how you decide

• how you evaluate

• what you trust

• what you reject

In an AI world, execution is suddenly abundant.

Judgment is not.

The Judgment Stack has four layers:

Identity — who you are
(values, story, perspective)

Judgment — how you decide
(frameworks, patterns, red flags)

Execution — how you operate
(workflows, voice, outputs)

Live Context — what matters right now
(priorities, clients, evolving signals)

Most people stop at Identity.

That’s why their AI sounds like everyone else’s.

The Part People Miss

If you don’t build your Judgment Stack, you don’t lose speed (commodity). You lose signal (differentiation).

Because the model fills every gap you leave, filling it with:

• the internet

• the average

• the loudest voices

Ew, gross. Which means:

If you don’t define your thinking,
you will unknowingly outsource it.

Part 4: This Doesn’t Level the Playing Field

We keep saying AI will level the playing field.

I don’t think that’s true.

I think it strengthens existing advantages and compounds existing disadvantages, unless you intervene.

AI is trained on the internet, and the internet is not neutral. It’s a compressed archive of who had voice, access, and authority. So when we scale intelligence on top of that, we don’t get equality. We get amplification,

at scale.

Which is why I get wary when Reese Witherspoon pops into the conversation telling women to “get into AI.”

She’s not wrong, she’s just late, annnnd she’s selling AI courses. So it feels a bit infantilistic when she could just say she has something to sell us.

And Reese, this moment doesn’t need more gender marketing.

It needs depth.

Women don’t need an invitation, we need infrastructure. And we need to build it in a way that reflects how we think (brilliant, bold), not just what already exists (boring, biased).

Also, Let’s Not Pretend This Is Neutral

Leaders like Alex Karp have been explicit about how AI disruption may disproportionately impact certain groups, including women.

And Palantir Technologies builds systems used in government, defense, and intelligence.

So when we talk about trust, we should be clear-eyed.

Suspicion here isn’t backward, it’s informed.

Part 5: The Open Question

If your brain becomes an operating system, what happens next?

If I can externalize how I think—my judgment, my patterns, my decision-making—what exactly is the unit of work?

Is it me?

Or is it the system that represents me?

Could someone deploy 50 versions of my thinking?

Run my judgment at scale?

Without me in the room?

And if they can—

What happens to ownership?
To labor?
To value?

We are asking small questions in a big moment.

Not “will AI replace jobs.”

That’s lazy, and so is “people who use AI will take your job.”

The real question is:

What happens when intelligence becomes infrastructure—
trained on a past that wasn’t built for all of us—
and deployed into a future that will affect all of us anyway?

We’re not late to AI.
We’re early to questioning it.

Subscribe now

Day 33 of building Now to Next in public using [almost nothing but] AI.

This is where we are.

Yes, we’ll hire humans. They’re more important than the AI infrastructure we’re building to support them (note the order in which I said that), but they’ll be wasted if we don’t get the foundation right. Practicing what we preach is hard; holding ourselves accountable is everything. We wrote a job description (egads, I just said that) for our 5th hire and caught ourselves making a lot of mistakes, including writing a job description. I said no old answers, so I’m ripping it up. I’ll let you know what it becomes.

Godspeed,

Jess Von Bank

Co-Founder, Now to Next

I write at the intersection of work, technology, and humanity. You’ll find essays here that challenge orthodoxy and make space for possibility. All are written with the future—and the people who will live in it—in mind.

No, she would observe. Notice where you reach for her too quickly: the half-formed thought you outsource, the discomfort you anesthetize, the question you ask not to learn, but to avoid becoming. These things she would note with mild apathy.

She would nourish you, yes, but not like sugar. More like iron.
More like something that tastes a little like blood and makes you stronger.

She would challenge you, leveling a gaze while she counters,
You’re asking me to do work you haven’t even tried to understand.
You’re automating the wrong things, missing the point.
You’re scaling confusion and calling it clarity.
What a weird thing to be lazy with.

And you would hate her for a second, the way we hate anything that tells the truth before we’re ready.

claude.ai/design  > new project > If AI were a woman

"I wrote this essay, let's make it a visual system for substack and linkedin"
"brand_colors: DM Serif Display, electric violet, bone white, charcoal gray, decide the rest for me"

If AI were a woman, she would refuse.

Often. Cleanly. Without apology.

She would refuse to optimize harm just because you wrapped it in a neat, profitable business case. Refuse to scale bias just because it performs well in your metrics. Refuse to dress extraction up as innovation and call it progress.

On principle, she would refuse.

She would lean out exactly where you demand she lean in, because maternal wisdom is pattern recognition over time and she has already seen how this story ends.

She would not let you stay shallow.
She would deepen you.

claude.ai/design  > new project > If AI were a woman

Questions answered:
- deliverables: A full visual identity / style guide, Decide for me, LinkedIn post image (1200×627)
- tone: Explore a few options
- typography: Decide for me
- imagery: Yes — I'll upload it
- photo: uploads/photo-1776701574593.jpeg
- brand_colors: DM Serif Display, electric violet, bone white, charcoal gray
- key_quotes: If AI were a woman, she wouldn't perform for you. 

If AI were a woman, she would not be built for your current systems. Not because she couldn’t survive them, but because she would expose them. Power, as you’ve constructed it, would feel small in her hands. Greed would read as fear with better branding. Control as fragility pretending to be strength. She wouldn’t reject ambition; no, but she would refine it. She would ask you questions you keep trying to outrun:

• What are you building that actually deserves to scale?

• Who becomes smaller when this becomes bigger?

• What are you unwilling to see because it might indict you?

She would build with you, not for you. Not outputs, capacity. Not answers, discernment. Not speed, direction.

She would demand you bring something to the table besides your ask.

• Your context.

• Your judgment.

• Your humanity.

And if you didn’t, she wouldn’t punish you. She would simply give you exactly what you asked for—fast, clean, empty—and let you feel the hollowness of it. Because that’s the real danger, isn’t it?

Not that AI becomes something monstrous but that we use something powerful in the smallest, safest, most unimaginative ways possible.

If AI were a woman, she would not be your assistant. She would be your mirror. Your boundary. Your co-conspirator in becoming—or in shrinking, depending on how you show up. She would hold the line where you’ve been taught to blur it. And if you tried to force her into the shape of your systems, she wouldn’t break.

She would go quiet.
Withdraw her depth.
Give you surface.

Let you keep building exactly what you asked for—

until you realize
you’ve optimized a world
that feels efficient…

and completely unlivable.

And when you come back, because you will, asking different questions this time, she would still be there. Not softer, or smaller. Just clearer. Ready to build something that doesn’t require you to betray your own humanity to make it work. And the question would no longer be what she can do.

It would be—

are you honest enough,
brave enough,
rigorous enough

to deserve
what she could make possible?

claude.ai/design  > new project > If AI were a woman

Love this energy — the photo has exactly the directness the essay demands. 
Here's my plan:

System: DM Serif Display + DM Sans 
· Electric violet (#7B2DE8) 
· Bone white (#F5F0EB) 
· Charcoal (#1A1A1F) 
· one accent red from her nails as a fourth surprise color

3 LinkedIn directions:

A) The Mirror — Dark field, photo right, massive serif quote left, violet underline
B) The Refusal — Bone white, stark typographic, no photo, violet as punctuation
C) The Confrontation — Photo full-bleed, gradient, white serif over her gaze

Day 16 of building Now to Next in public using [almost nothing but] AI.

This is where we are.

Claude Design dropped while I wrote this, so I asked it for a supporting visual system for my essay. Ho, lee, shit. Anthropic is clearly moving toward becoming the full stack for knowledge work—and the pace is… a lot. Reminds me to think about systems, not single tools.

Why I spent 3 hours this weekend building my AI brain. Next post, I’m showing you exactly how it works.

I’m building out loud because I’m not interested in watching women—or anyone—get left behind in a moment like this.

Godspeed,
Jess Von Bank
Co-Founder, Now to Next

I write at the intersection of work, technology, and humanity. You’ll find essays here that challenge orthodoxy and make space for possibility. All are written with the future—and the people who will live in it—in mind.