She did not choose Oz. She did not train for it or put it on a roadmap. A storm picked up the house with her still inside it and set it down somewhere the rules were different. And then everyone she met asked her what her plan was.

One of my leaders told me we were moving through the five stages of grief.

Denial. Anger. Bargaining. Depression. Acceptance. I have watched every one of those stages show up in a meeting since.

This task force hit home because this transformation is in my house. Product, engineering, design, research, learning technology, communications, Salesforce, emerging technology. I have written about workstreams from the outside. This one I stand in every day, and proximity is not clarity. Sometimes it is just a better seat for watching something hard.

It is also the one that taught me the thing I least expected, which is that the hardest part of this has almost nothing to do with the disruptive technology.

The Tornado Did Not Take Requests

We are not in Kansas.

For a long stretch, the loudest thing in the room was disbelief. Not resistance exactly. Something closer to “surely not us”. Surely not this craft. Surely not a discipline that took fifteen years to get good at. Other teams, sure. Other companies. Not us.

I understand it. That is what expertise feels like from the inside.

The disruption did not come for the skills first. It came for the seams. The tidy lanes we spent a decade drawing. Product writes the spec. Content writes the words. Design crafts the thing. Engineering builds the thing. Research validates the thing. Handoff, handoff, handoff, ship. That system was not an accident. We built it on purpose. We became a machine. As we grew, we also slowed down: more checks, more approvals, more workflows.

Now one person with the right tools can stand in multiple lanes at once, and the handoff that used to be a safety net is starting to look like a delay.

You cannot reorganize grief. You have to let people move through it.

The Person Behind the Curtain

When I stand in front of my team and describe what is coming, I am seeing a wizard. Real magic. A genuine unlock. And some days I know, with total clarity, that what they are seeing is a woman in a costume with a microphone, pointing at a curtain and insisting there is something behind it.

That is the actual risk of leading this. Not that you are wrong. That you are early, and early looks exactly like pretending.

The wizard in that story had no power except that people believed he did. That is the accusation sitting quietly under these conversations. You are selling us something you have not built.

Fair. So let me tell you what I think is behind the curtain, and what it costs to find out.

They Gave the Tin Man a Clock

At the end, the wizard does not give the Tin Man a heart. He gives him a clock. A ticking heart-shaped watch, so he has proof of a thing he had been doing for the entire film.

The heart is the hardest one. It always is.

What do you say to an engineer who wants to write every line themselves? Not out of ego. Because writing it is how you know how to fix it when it breaks.

I am not asking anyone to know less. I am asking them to know something that will not hold still. What these tools can do this month, where they quietly fall apart, what context they need to be worth anything, and how to improve the system so the next person gets a better version than you did. That knowledge sometimes has a shelf life measured in days. Then spend what is left on the problems that are genuinely hard, which is where I wanted their expertise all along.

What do you say to a designer when you hand templates and a design system to the rest of the organization so they can make their own deck, their own graphic? On paper it is a gift. In practice you are telling someone the work they have been buried under for years is now self-serve.

Except. Was that the work? Was any designer ever chasing another person’s presentation?

Or did they want the time to think. To build a brand nobody has seen. To sit inside a complicated experience long enough to make it feel effortless. To be a strategic leader instead of an asset factory.

The floor of design has come up. So has the ceiling. Getting to the ceiling is the whole job now, and we have not cleared the time for it yet.

The grief is real.

They Gave the Lion a Medal

The wizard did not hand the Lion courage either. He handed him a medal. For courage the Lion had already spent the whole movie demonstrating, badly, while terrified.

I want to be careful here, because comfort has a bad reputation it did not earn. Comfort is not laziness. Comfort is the reward. You spent years building a way of working that holds up under a bad quarter, that you can execute on your worst day and still be proud of. Nobody should have to apologize for wanting to stand on it.

The ask right now is to step off it anyway.

To try something in the open while you are still bad at it. To let people watch you be a beginner at the exact thing you are supposed to be the authority on. To risk a way of working you cannot yet guarantee, knowing some of it will fail, and knowing some of it will contradict a thing you built.

That is not a skills problem. We spend years learning to walk into a room as the person with the answer. Now the useful thing to be is the person with the question. I cannot ask for that and not do it. If I am not visibly wrong in front of my team on a regular basis, then I am not exercising courage.

So we changed how we meet. My broader leadership team, not just product and engineering, now sits down every week on this and nothing else. Some weeks it normalizes. Some weeks it storms right back. We bring in builders to show their work, not to be inspired, but to get our imaginations unstuck watching people we respect do what we assumed was not yet possible.

Radical candor. Radical honesty. Radical acceptance. Acceptance is the hardest one.

My leaders came with the real questions and I published every one rather than smoothing them over. Why do we need to go faster. What happens to specialists. Are you defining skill as a workflow or knowledge. Where does the bandwidth come from. What happens to junior talent. Those are not obstacles in the way of the vision. They are the vision, being built with the people who have to live in it.

We also review real case studies we are hearing, so we can keep what makes sense, and throw out what doesn’t. I would much rather disagree now than find out six months from now that we were never actually agreed.

They Gave the Scarecrow a Diploma

The Scarecrow does not get a brain. He gets a diploma. A credential for a thing that was never missing in the first place, from a man who could not give him one anyway.

Then there is judgment. The brain.

Here is what I believe. Software is going to be cheap. Everyone is a coder now, and we should stop pretending that is a provocative statement instead of a description of Tuesday. The floor has come up under writing, under design, under code, and the gap is not closing back up.

Which means the scarce thing was never the building. The scarce thing is knowing what is worth building, when it is actually done, and what good even means here.

That asks you to spend more time thinking and less time producing, which is deeply uncomfortable for people whose entire career reward system was shipping.

So, why are we building anything repeatable?

If the work is repeatable, someone else can now do it, which means it no longer differentiates us, which means we have already fallen behind. If somebody else can produce our work, we are not doing our job. Our job, as specialists, just got harder than ever, and it is not survivable without systems built underneath us to hold quality at scale.

We are not trying to do the same work with fewer people. We are trying to serve a far larger surface of impact with the people we have.

Nobody Is Laying Brick

And yet.

We are absolutely faster than we were. Individually, dramatically faster, more creative, more self aware. Solutions in weeks that used to take months. We are learning what scrappy means again, a skill our org had quietly lost in some ways.

But we have not unlocked the yellow brick road.

Individual speed is not a road. A road is a system. A road is what lets the next person travel without discovering the route from scratch, and we have not built ours yet. We have not built the system that would make us less busy. We are sprinting through a field and calling it progress because sprinting is genuinely faster than walking used to be.

I have also noticed that parts of the tech team are becoming the enablement team. Not the only builders, but the people who make it safe and possible for everyone else to build. An enablement plan. Maturity models that give people room to explore while being honest about where oversight is not optional. Even rebuilding our design system as infrastructure for building at scale.

Everyone else raises the floor. That is exactly what raises our ceiling.

Which means letting go of work that no longer serves us, even when someone else picking it up feels like something being taken. One of my favorite articles is give away your legos, and we are living it right now.

“Why Didn’t You Tell Her Before?”

At the end of the film, Glinda tells Dorothy she had the power to go home the entire time. The Scarecrow, who has walked the whole road and been set on fire for the trouble, asks the only reasonable question. Why didn’t you tell her?

Glinda’s answer is the best line in the movie.

Here is what I got wrong at the start. I scoped this workstream as engineering. Agentic engineering. Build faster.

Engineering turned out to be one link in a much longer chain. Nobody here delivers value alone. Product cannot create delight without design. Design cannot ship it without engineering. None of it reaches a student without integration into the LMS, without communications, without the systems underneath all of it. Change how one part of that chain works and you have changed every part of it. Not a harder transformation than the other task force groups. A wider one.

Which is why the variable that decides who thrives has almost nothing to do with AI.

It is not tool fluency. The people moving through this well are the people who embrace new things. Those skills include:

• Lifelong learning. Any one skill has a shorter shelf life than it used to, and that is not a threat. It just means curiosity is the thing worth protecting. Stay a little bit of a beginner on purpose.

• Navigating ambiguity. Nobody has the map. You do not need certainty to take the next step. Decide with what you have, stay open to being corrected, and keep moving.

• Resilience. It will not work the first time. Or the third. That is the process, not a verdict on you. Give yourself the patience you would give someone you were teaching.

• Growth mindset. We are all trained to ask how we make this ten percent better. The question worth sitting with now is what it would take to make it 10x better. That is not a referendum on what you built. It is the invitation to build the next thing.

• Communication. When the handoffs stop being clean, communication becomes the machinery. Right now we are all quietly being brilliant in separate rooms. Transformation requires more communication, not less. More visibility into what you are trying. More vulnerability about what is not working yet. More alignment, said out loud, more than once.

Nobody in this building is going to believe what is possible because I said it at an all hands. They are going to believe it the first time they do something they did not think they could do.

The Door Only Opens One Way

Does it actually matter that the wizard was not real? Look at what he actually handed out. A clock. A medal. A diploma. Props. Every single one of them proof of something the person already had before they ever walked through his door. The magic was never in the man behind the curtain. It was in the fact that they walked.

That is the whole thing. It is not up to the technology. It is up to us.

And it does not work the way we want it to. We want to see it first and then believe it. That is not the order. You walk, and then you see it, which means somebody has to leave comfortable right at the moment comfortable was finally within reach. That is the ask I am making of my leaders, and of myself.

None of this requires changing the people doing the work. It might just mean the work finally becomes the work we signed up for. We are learning to trust each other on different terms, to communicate on different terms, and our collaborative intelligence is going up because of it, not down.

One more thing about that movie.

Kansas is shot in black and white. Sepia, if you want to be precise. Flat and gray and full of people telling Dorothy to mind her own business, and she cannot get out of there fast enough. Then a door opens and the whole world is in color, and she spends the rest of the film trying to get back to the gray.

I understand her. Gray is legible. Gray is where you already know exactly what you are good at.

Not everyone on this team is standing out here in the color yet. But we have stopped disagreeing about whether there is a road. We are communicating about where it goes and what to build it out of, and that is a completely different conversation than the one we were having a few months ago.

They have seen the color. The road is not finished. We are laying brick while standing on it.

And some days my team still sees the costume. I have stopped trying to fix that, because here is what I finally understand about the story.

Nobody was changed by the wizard. They were changed by the walk. The heart, the courage, the brain, every bit of it came from the journey. The man at the end just handed them something to hold.

The road was the wizard the whole time.

So I am not asking anyone to believe me. I am asking you to come walk. The road will do what I cannot.

Five workstreams: Success Coaching, Course Building, Agentic Engineering, Content Operations, PMO Automation. This one is about coaching. I am writing it as a teaching piece. If you run a coaching center, a call center, an advising team, or any function where humans talk to humans at volume, my hope is that you can read this and diagnose your own version of the problem.

TLDR

• The core problem was a ratio. Coaches spend approximately one third of their working time in direct connection with students. The remainder is consumed by the operational work surrounding that connection.

• Our CRM is a product and our coaches are the users. Which makes our Salesforce team a product team.

• Loyalty hides friction. People adapt quietly instead of complaining. The cost is still real, and in our case it was two thirds of a day.

• Fix your knowledge base first. The least exciting thing on this list beat every model decision we made.

• Ship it while it is bad. Live mediocrity teaches you things faster than waiting to get it perfect.

• Your evaluation expired. Anything you rejected more than a year ago needs retesting, and your organization’s memory of that rejection is usually the real obstacle.

• Trust is the constraint, not capability. Adoption is behavior change, and behavior change is slow if you roll out without thorough testing and clear expectations.

What Is Coaching?

The university has different groups of specialists. Financial aid counselors are subject matter experts on student finances. The admissions team knows admissions criteria, housing, and recruitment. Transfer specialists know transfer applicability and review. Advisors support students with degree planning. And so on.

Each of those groups is excellent at their piece. Each of them owns one part of the student journey.

Coaching is the function that lives alongside a student from interest to graduation. We have hundreds of coaches and leads supporting our 100,000+ online students.

Here is the structural principle underneath both. Institutions are organized by function. Students experience their lives by problem. A prospective student arrives with a dozen questions living in nine departments, at the moment their life is hardest and the financial stakes are highest.

Coaching is the translation layer between those two facts.

That is why I call it the heart of this organization. Take the sentiment out and it is still true operationally. It is the only role that holds the entire student.

Find Your Ratio

Coaches spend approximately one-third of their time in direct contact with students, measured across all channels we run, including voice, SMS, chat, and email.

The remaining two-thirds are absorbed by the operational work surrounding that connection.

Two-thirds of the day, from people whose entire reason for being here is the conversation, was spent on everything except the conversation.

Accurate diagnosis matters here, because the intuitive explanation is wrong. The systems are not deficient. Each tool a coach uses was implemented by someone addressing a real constraint at the moment it appeared. The CRM replaced spreadsheets that no longer scaled. The channel mix reflects where students actually engage. Third-party platforms were adopted because internal development would have required time the organization did not have.

Individually, every layer represents sound work under pressure. Collectively, those layers are the reason the organization can serve its current student population at all.

They are also the reason a coach must hold ten variables in working memory to resolve a single question.

The mechanism is worth naming precisely. Accumulated complexity does not arrive as a decision. It arrives incrementally, below the threshold at which anyone conducts a review. That is a normal consequence of growth, not a failure of governance.

Then apply scale. That volume does not create infrastructure pressure. It creates pressure on the people who engage students one at a time. And a ratio problem is not resolvable through headcount, because adding staff to an unimproved ratio increases navigation capacity, not coaching capacity.

Our CRM is a product. Our coaches are the users.

Our CRM is a product. Our coaches are its users. That makes our Salesforce team a product team, accountable for user outcomes rather than configuration throughput.

Configuration work became design work. Support tickets became research inputs. The acceptance criterion shifted from whether a function performs correctly to whether it performs well for a coach on their eleventh conversation late in the day.

The broader observation is worth stating for other organizations. Internal tooling is rarely held to a customer-facing standard, and not by deliberate choice. It is because staff are loyal. They adapt. They build quiet workarounds and never mention it, because mentioning it feels like complaining.

Loyalty renders friction invisible. Invisible friction remains a cost. In our case it accounted for two thirds of a working day.

A Note on Buying Versus Building

Configuring inside somebody else’s platform is a different discipline than greenfield development. You inherit their roadmap. You wait on their release cycle. Your innovation ceiling is partly set by a company you do not control.

We evaluated Agentforce approximately two years ago and determined it was not ready for our requirements. That assessment is no longer accurate. The platform advanced more in the past six months than it did the two years preceding.

Any negative assessment older than twelve months should be considered unreliable, and the organizational memory of that assessment is usually the obstacle to adoption. Put re-evaluation on the calendar the way you schedule security reviews. Twice a year, with a named owner. Assume every negative verdict older than twelve months is wrong until re-proven.

We Come With Receipts

Student chatbot

• Deployed on the ASU Online site several years ago. Initial performance was mediocre and required substantial fine tuning and significant improvement to the underlying knowledge base.

• We kept it live anyway, and that is why it works now. A live mediocre system generates the one thing a conference room debate never will: evidence. Every conversation told us what students were actually asking and whether our content answered it. Frequently it did not, and we would not have known.

• Today it maintains a persistent personalized profile for returning students and nudges proactively. We see over 90% deflection on questions that used to route straight to a coach and were identical every time. Deflection here does not mean students got less help. It means the repetitive tier one question got answered instantly at eleven at night, which provide real-time support for students, and gave the coach back time for the conversation only a human can have.

• Planned expansion includes the current student experience, which was recently launched on our website, where we are repeating the same evidence collection.
  

Conversation transcription and analysis

• Through our telephony provider NiCE, coach and student conversations are transcribed, analyzed, and summarized directly into the CRM.

• There are two benefits and they operate at different levels.

  • First order: presence. Think of every meeting you have sat in typing furiously and absorbing nothing. That was the coaching call. A coach can now be fully in the conversation during what is often the hardest conversation of that person’s week.

  • Second order: a data layer we did not have. We used to say we talked to ten students and here is what we heard. We can now aggregate themes across tens and hundreds of thousands of conversations, drop into the actual quotes, and locate where in the journey a hesitation appeared, because it is tied to the CRM record. Every objection and confusion point in the funnel, at population scale, in students’ own words.

Pre- and post-call automation

• AI-generated summaries of existing student context are available before the conversation. Following the call, a transcript analysis identifies previously attended institutions. This triggers an LLM to perform a web search to find where the student can request each institution’s transcripts from and creates an outbound communication for the coach to review. This augments the coach by reducing the need for them to look this up by hand.

Degree comparisons

• Coaches now have the ability to compare multiple programs in real time through Agentforce, while on a call, without needing to look up each knowledge base entry and determine the differences between similar or different programs.
  

A knowledge base rebuilt for machines

• Least glamorous item here and the first thing I would tell you to do. Our accuracy improvements traced back to this more than to any model or vendor decision.

• The recommendation is to design knowledge architecture explicitly for language model retrieval, and to staff that work with content and SEO practitioners. Structuring information for machine discovery and reliability is precisely the discipline those teams have practiced on the web for years, and the capability typically already exists internally.

• We additionally reduced our knowledge card volume by more than half, improving maintainability and human navigability.

Call review at scale for leads

The best ideas are generated by the coaching center themselves; this is an example

• Leads used to listen to calls one at a time to prepare for one on ones, which capped how much they could observe. They now build criteria templates, review at volume, and arrive with concrete examples of what a coach did well and where the opportunity sits.

• Coaching quality improved measurably during the pilot. Notably, the improvement was attributable to feedback quality rather than feedback volume. Reducing the retrieval burden allowed leads to analyze patterns rather than locate examples.

Just Because You Can, Does it Mean You Should?

The knowledge base interface remains in service. Coaches continue to search directly, and we are still assessing whether full generative retrieval is the appropriate behavior change and what it costs at volume. In some cases the existing interface is faster and less expensive. This is a legitimate tradeoff to evaluate on the numbers. One we are still exploring.

Adoption is the actual constraint. Not capability. Trust.

A coach who got a wrong answer in March remembers it in July, and they are right to. Accuracy is not uniformly there yet. Asking people to trust a system that has burned them is not a strategy, it is a wish. You rebuild it with alignment, communication, and a volume of patience that feels deeply unproductive while you are spending it.

Budget for that. No business case has a line item for it and every timeline depends on it.

We are also still redesigning the coaching model itself with center leadership. More intentional proactive outreach across SMS, voice, and chat, with nudges that persist with a student rather than resetting every time the channel changes.

10x The Idea

The most valuable outcome of this workstream was the pressure it applied to our technology teams.

Coaching forced a question that is uncomfortable to answer honestly. What is the 10x idea here, rather than the 10% improvement. Technology organizations tend to optimize systems they have already built well, because that terrain is familiar and the work feels productive.

The discipline required is to identify the point of greatest need and redesign the process rather than the interface.

Still Loading Issue 6 was me worrying about speed outrunning judgment. This is the other half of the same argument. The human in the loop is not a compliance checkbox at the end of a workflow. Here, the human in the loop is a coach on a call with someone deciding whether they can afford to finish their degree.

Everything we automated exists to protect that call.

Give Them the Job Back

Ask a coach why they took this job. Nobody has ever answered with a system.

They answer with a student. The one who almost did not enroll. The one who went quiet for three weeks and came back because somebody noticed.

Thank you to Nancy Cervasio, Nicolette Miller, Allison Jesson, and Kaleb Anderson, who led their teams through four months of experiments (and many more to come) that did not always work on the first try and stayed curious anyway. You do not transform anything without leaders willing to lean in.

Aly Rassel is a project manager at EdPlus and part of the PMO Automation workstream I wrote about in Issue 2. She was a high school history teacher who walked into this place never having opened Jira. Five years later, she’s the person we asked to aggressively test whether AI could take the busywork out of project management without taking the human out of it.

I asked her to write about what that actually felt like. She delivered.

Here’s Aly.

From Chalkboards to Kanban Boards

My project management career started about 5 years ago with EdPlus at Arizona State University. I had no experience in a corporate environment, had never used Slack, never heard of Jira. I became a high school history teacher right out of college and never looked back. But in 2020, COVID hit, and I had my first child. It was time for a change. I saw a project coordinator role at ASU, my alma mater, and thought it was the perfect opportunity to remain close to my passion for education while making an impact in a new way. I quickly learned new tools, how work was tracked, the software, and design development lifecycles. My job was managing intake, cross-functional deliverables, and documentation. I took notes and scheduled meetings. I followed up on assigned tasks, reached out to stakeholders, and kept leadership in the loop. In a matrixed organization with 400+ employees, these roles are essential to keeping things on track and under budget.

Is This Good, or Is This Coming for My Job?

As I gained experience and was promoted to project manager, my role shifted from team sprint management to longer lifecycles and resource planning with leadership teams. And because I was managing large projects for the Learning and Emerging Technology team, I had a front row seat to how AI was transforming our work. I learned the capabilities. I learned the lingo. I started to pick up some of the backend functionality. The team empowered me to experiment alongside them and test out the tools. I was a de facto beta tester for all the automations that they built. That’s when I started to reflect on how these tools would change my work (and my life): Is this technology good, or is this coming for my job? What is the cost of this technology: diminished critical thinking, over-reliance, and environmental impact? But surely, there would always be a need for a human-in-the-loop (right?), so how could I use this technology to enhance my work rather than replace it?

The PM Task Force: Automating the Busywork

In early 2026, the AI train was moving whether I was on board or not, so I hopped on and kept testing, iterating, and seeing increased productivity and efficiency. EdPlus recognized the transformation potential across the org and created several AI task forces, led by Amanda Gulley, our Chief Product and Experience Officer. Since I was so close to the existing automation/AI teams, I was asked to join the task force dedicated to enhancing the Project Management Office (PMO) through AI automation.

This included rethinking status reports, artifact generation, meeting preparation and recap workflows, cross-team communication norms, and note-taking and automated action items. I was asked to aggressively test automations, AI copilots, and workflow redesigns to shift PM capacity toward higher value orchestration and strategic enablement. To do this, I was granted access to AI and automation platforms, including a meeting transcription and AI assistant, connected to our communication channels. The door was now open to see how I could challenge existing processes, eliminate redundant layers, and make visible where PM effort is ripe for automation and AI integration.

From Documenting the Work to Shaping It

For a few months, I spent most of my time on the PM Taskforce. This was a blessing and a curse. I had transitioned from the Learning and Emerging Technology team to the Salesforce team just 3 months prior. While I was still onboarding to a new team and new projects, I had a unique opportunity to leverage automations and workflow redesigns to benefit myself and my new team. I worked with the Director of Learning and Emerging Technology, Auryan Ratliff, to build a PMO plugin in Claude. The plugin would leverage meeting notes in Granola for the knowledge base and provide a dashboard in Claude that lets me connect to a Jira board, create and update Jira tickets based on recent meetings, and draft a Slack recap for each meeting. Because the human-in-the-loop was so critical to me, each step within the plugin requires human review and allows for edits before pushing the update.

Meeting recaps, complete with discussion topics, decisions, and action items, were posted in Slack within minutes rather than days. Jira tickets were being created with more technical detail and requirements. Project hub updates were being automated: meeting notes were posted, dates were logged, and action items were checked off. Product feedback sessions were being converted into tangible spreadsheets that could be prioritized and monitored. My role quickly changed from documenting the work to helping shape it: scope, timelines, dependencies, pulling leaders up to the big picture. I am now in meetings instead of typing them. I can serve as a consultant and thought partner on my projects. I’m now managing processes and systems rather than spending time with transcription, ticket creation, and Slack updates.

I'm Not an AI Evangelist.
But It Gave Me My Job Back.

I’m not sure I’ll ever call myself an AI evangelist. The concerns are legitimate and important. What I’ve learned over the past few months is simple: the routine work is being done faster and more efficiently, leaving room for me to challenge assumptions, sharpen goals, and bring a fresh perspective. The hours I used to spend on transcription, ticket-writing, and status updates now go toward listening more closely, asking better questions, and helping my team think through the hard parts. I could not have done any of this alone; it happened because a team was generous enough to let me experiment alongside them.

“Where do I even start.”

Here is the honest problem. These tools change faster than I can write the instructions for them. What I tell you today has a shelf life of about a Tuesday.

So this is not the definitive guide. It is where I started, and where I would start now, knowing what I know.

TLDR

• Stop shopping for the perfect tool. Most of them do the same things.

• Pick one professional tool you already have access to and run couchto5k.ai. A few minutes a day, 30-days.

• Pick one personal project and bring it into Claude Cowork on a free or paid Pro account. 30-days. It’s worth the $20 I promise.

• Use one or two tools daily, not ten occasionally. Consistency beats collection.

• Learn prompting and context before anything fancy. Skip skills, plugins, automations, and live artifacts for now.

• Start every tool by telling it who you are. It gets exponentially better the more it knows about you.

First, Stop Shopping

Nobody tells you this because it does not sell a subscription. Most of these tools do the same things.

Ask ten people which one to use and you get ten answers, each delivered with total confidence. The Google person swears by Gemini. The engineer will not stop talking about Copilot. The operator lives in ChatGPT or Claude Cowork. They are all right. They are all describing their own job back to you.

Some models are genuinely better at certain things, and I will tell you which ones I reach for. But the gap between the right tool and the wrong one is almost nothing next to the gap between picking one and actually using it.

How to think about the categories before you pick:

• Generative tools answer, draft, and make things when you ask. Chat interfaces, image generators. This is what most people already know.

• Agentic tools take actions across steps and surfaces. Claude Code, Claude Cowork, Codex. This is the part most people have not felt yet, and it is the part worth getting comfortable with.

• Embedded tools live inside software you already use. Gemini in Google Docs, Figma’s built-in chat, Agentforce’s agent builder, Slack’s new Slackbot integration. Lowest friction if you already work there.

You do not have to learn all at once. You have to pick one tool and start.

Sorted by the Job, Not the Logo

You do not wake up wanting an LLM. You wake up needing a graphic by noon, or a one-page executive report that must summarize 10 pages of detailed notes, or finally remembering what was decided three Slack channels ago. So here is the kit based on what you are actually trying to do.

Images and graphics

• ChatGPT. My go-to, and about the only thing I open it for. Blog and social graphics, Slack announcement images, slide visuals, even style guides.

• Gemini (nano banana). Close second. And honestly, just as good. I have access to both, but today ChatGPT is slightly better.

• The move that matters more than the tool: give it an example image, not a description. Find the thing you love (inspiration image) and hand it over. Inspiration beats interpretation every time.

Building things (reports, dashboards, websites, internal tools)

• Claude (Cowork or Code) and Codex. Neck and neck, both excellent. You do not need to code. You talk to it in plain (natural) language.

• Tips: hand it a brand name for style direction. Apple and Airbnb are my defaults, public design systems the model knows well, and they pull clean and minimal. Or feed it your own brand standards and link the page where your colors and type live, so the output belongs to you. Again, examples (context) are key when wanting something that looks good. Show it what good looks like.

Design work

• Claude Design and Google AI Studio. My favorites for honing one specific component before carrying it into other files.

• Figma Make, plus the new chat built into Figma. My teams use these, and the in-Figma chat has impressed me. Again, if you aren’t a huge Figma user, then skip this one.

• Claude Design for presentations. Best slide and data visualization I have used. You cannot tell my decks from my design team’s. Hand it a template, a Figma file, or a design system, and it stays on brand.

• The catch: Claude Design burns credits fast. On a Pro account, it drains quickly. I would not bother unless you are paying for Max yourself, which is $200 a month. I do not love saying that number out loud. It is the truth anyway. So sometimes your choices will come down to cost.

Meeting notes

• Granola. One of the most transformative tools I use. You stop scrambling to capture notes and start being present. It draws on past meetings for context and captures action items. This is one of the few tools that does not record audio or video of your meeting. Again, try for free, but make sure you do not record meetings that includes sensitive data.

• Zoom AI Companion. If your meetings already live in Zoom, the built-in companion will summarize the call and surface action items without adding another tool to your stack. Lowest friction if you are already there. Not my favorite tool, but you can download the transcript and also add it to a folder so an LLM can access it. I find this way better than the native Zoom AI features, but it requires you to be organized and take an extra step to capture the transcript.

Voice to text

• Willow. It does not just transcribe, it takes my rambling and articulates it cleanly, stripping the “ums” and false starts better than any built-in voice tool. It can now act on Slack or a Google Doc on whatever surface you are working in. Cheap, easy, permanent fixture. It also just released a new feature called “scribe”, check it out for free (or $10/month).

• Note: There are many of these out there; again, this is all based on preference. I am sharing my toolkit with you, but you may have your own!

Slack/Communication

• First, I organized my team's Slack updates by having everyone standardize weekly or biweekly project updates in Slack Canvas.

• Slack canvas + Slackbot for keeping project updates in one place, you can pull through any AI tool or a weekly summary bot, instead of crawling through twenty channels. It knows a startling amount about how we work. You can even ask it the hard version: what could I do better as a leader, how is my communication landing? It will coach you if you let it.

• Example prompt for Slackbot

  • Finding & summarizing information

    • “Catch me up on what’s happening in [channel]”

    • “What did [person] say about [topic] this week?”

    • “Find all messages about [topic] from the past month”

    
    Tasks & automation

    • “Remind me every Monday morning to check my [x]”

    • “Set up a weekly digest of [channel] every Friday at noon”

    
    Build your “About Me” canvas

    • “Create a canvas about my role as [role] — include what I do, what I own, how to work with me, and what I care about”

    • “Write up a canvas that introduces me and my team to new colleagues”

    • “Make a ‘Working with [name] guide as a canvas — my communication style, priorities, and how to get decisions from me”

The true coworker

This is the part most people have not felt, and it is the part that changes how you think about these tools.

• Generative AI answers when you ask. Agentic AI goes and does. It takes a goal, breaks it into steps, and works across surfaces without you having to hold its hand through every click.

  • Claude (through Claude Cowork and Claude Code) and Codex (OpenAI) are the agentic tools I reach for. Picture the difference this way. A generative tool drafts the email you describe. An agentic tool reads your inbox, finds the three threads that need a reply, drafts all three, and you choose to go to your drafts to review before you click send.

  • At full strength, that means reading your email, calendar, AI meeting notes, local folders, Google Docs, Slack, and handing you a real prep on your day. It not only tells you what your action items were from last week, it already began working on them and you have drafts or code ready for your review.

  • The catch is integration, aka context. To really work, they have to plug into your systems, and at most organizations, ours included, those integrations are not yet cleared through security. You can feel the agentic behavior today on smaller, self-contained tasks. The fully wired-in version is still loading.

  • Gemini in Google Drive and Docs is the exception. If you live in Google, you can start drafting in real time today.

The Skill Nobody Sells You

Here is what I keep watching happen. People use these tools without understanding the cost of using them. Almost all of it comes down to one skill nobody teaches.

Prompting. And right behind it, context.

• This matters more than the tool you pick.

• When your context window fills past 30 or 40%, especially while coding, hallucinations creep in. Keep it lean.

• Most people never learn the fundamentals because they got comfortable on day one and never leveled up. Same habits, shrinking returns.

We have recently been pointing our teams to Couch to 5K for AI. Exactly what it sounds like. You do not sprint a marathon on day one. A few minutes a day for thirty days, walking you through what different prompts do, how context shifts the output, and how to tell whether a prompt is any good.

What to Skip for Now

You will hear about a hundred advanced features. Ignore them until the basics are muscle memory.

• Skip skills. Not yet.

• Skip plugins. Not yet.

• Skip live artifacts. Not yet.

• Do focus on getting comfortable with one or two tools and feeling the difference between generative tools and agentic ones like Claude Cowork, Claude Code, and Codex.

Your Actual Assignment

The truth about every tool above is the same. They do not work because you collected them. They work because you used one consistently long enough for it to learn you. So, two things.

1. For work. Pick a tool you already have access to. The approved one. Do not wait for the perfect setup. If none are approved, do this as a personal project on your personal device, using your personal accounts.

• Go to couchto5k.ai.

• A few minutes a day. Thirty days.

2. For you. Find one personal project and bring it into Claude Cowork. Free or Pro, it does not matter.

• Thirty days just to see what it can make and what it can dig up.

• Example: I had it go through my personal email and pull every summer camp receipt into a spreadsheet so I finally knew which kid was where, which week. I had a full spreadsheet broken out by kid, which camp each week, location, hours, cost, and notes on what they need to bring. I was organized in minutes, which would have taken me hours.

• Don’t know how to do something? We got you. Force the curiosity. Ask it directly, give it your role, a problem, your task for the day, and ask it to work with you on the solution, or ideas of where it can see efficiencies or creativity in your idea. Ask it what you would ask an expert, and when the answer turns too technical, tell it to explain it like you are in fourth grade.

Before any of that, do the thing almost everyone skips. Have a real conversation with it about you.

• Who you are. What you do. What is hard about your job, or your life. What you actually enjoy.

• Then ask how it could help you be more present, more efficient, more whatever you are reaching for.

• You can even give your agent a persona, or a personality. I ask mine for “real-talk”. I enjoy a LLM with some personality and fire, so I gave mine some spice.

Because this is the whole secret, the one the Couch to 5K teaches you by the end. These tools get exponentially better the more they know about you. You are not using a search engine. You are training a collaborator. It cannot collaborate with a stranger.

Pick one or two. Show up daily. Let it learn you. Watch what happens at day thirty.

Just Get In

This is where I would start. Plenty of smart people will hand you a different list and tell you I am wrong about half of it. Maybe they are. There are good articles out there that will round out everything I just said.

Start. Pick the thing. Show up tomorrow. Then show up the day after that.

The floor is rising fast. The only wrong move left is standing perfectly dry at the edge of the pool, still googling how to swim.

Go get in.

I have been hardest on the design team. Harder than I’ve been on anyone. And it’s because I am one of them.

You hold the thing you came from to a higher standard. You see the gap between good and great faster than anyone in the room because you spent a career living in that gap. That’s the bias. I’m naming it up front because the rest of this only makes sense if you know which side of the glass I’m standing on.

Steal Like an Artist

A prototype landed on my screen last week. Built by someone who has never sat next to a designer, never opened Figma, would not know a baseline grid if it introduced itself. Product, maybe engineering. Couldn’t tell. And it was good.

Not great. Not the kind of thing you screenshot and send to a friend. But good. And six months ago, it would have needed us. That’s the part that stings. It doesn’t anymore.

That’s the whole year in one screen. Bad design is becoming mediocre. Mediocre is becoming good. The floor is rising and it is rising fast. Anyone can grab an animation off Dribbble, a layout off Pinterest, a full design system off a brand they admire, feed it to a model, and get something back that holds up.

I’ve done it. You’ve done it. We have all quietly done it. Because that is what designers have always done. You find three things you love and you melt them down into something that’s yours. Steal like an artist. We practically trademarked it.

The plot twist is that everyone can do the melting now. The secret recipe is printed on the box.

So here is the question that kept me up at a genuinely unreasonable hour. If our craft was the thing that set us apart, and the craft now ships with a free trial, what exactly are we for.

“Yup, That Was Made By AI”

You know the look. Every doc built on the same skill comes out wearing the same outfit. Same spacing, same gentle gradient, same rounded corners that are trying so hard to seem friendly. You can spot a vibe coded interface from across a crowded room, and lately the room is very crowded.

The models hand everyone the same answer because that is the entire job of a model. Give the most probable response. Probable is just another word for what everyone else is also getting.

For a team whose whole identity was not looking like everyone else, this lands somewhere between insulting and existential. The work was the fingerprint. You could pick our designs out of a lineup with the lights off. So when the lineup starts looking like a row of identical twins, you feel it in your stomach a full beat before your brain catches up with the words for it.

I felt it as the Deputy of AI and Automation. I felt it a lot harder as a designer.

Hold My Figma

Now let me tell you about a transfer tool.

A transfer tool is about as unglamorous as software gets. Upload a doc, stare at a wall of course numbers that mean nothing to you, and good luck decoding any of it. Nobody has ever written a poem about a transfer tool. Nobody has cried at one. It is the software equivalent of the DMV.

One of our designers made people feel something at the DMV.

Yes, it was a redesign. Yes, it was a big swing. But that is not the part that got me. It was the animation. The thought behind the celebratory moments. The little bursts of joy someone clearly agonized over, assets I had never seen anywhere before, nothing off a shelf, nothing off a model. It was crafted with deep care, and you could feel it. The kind of subtlety you only reach once you have stopped pushing pixels and started thinking about how an actual human moves through an actual moment, and how you want them to feel when they get there.

And then, and this is the part that got me, they kept going. Five steps past where the brief ended. Into territory they had never once had time to explore, because for years they were buried under the flyer, the deck, the asset somebody needed by five o’clock yesterday.

That was the exact moment my little identity crisis packed its bags and left.

Because here is what I finally got. Everyone can copy what we already made. Fine. Let them. Frame it, even. They can rebuild last quarter’s work all day long. But while they are busy catching up to where we were, we are already somewhere they have not pictured yet. We do not stay static. We have never once stayed static. The craft was never the artifact sitting in the folder. The craft was the refusal to stop pushing it further.

As Elsa Says, Let It Go, It Does Not Serve Us

Here is a truth that should be on a poster somewhere. Nobody wants to wait for a loading screen. Nobody wants to fill out a form. Nobody has ever, in the history of being a person, wanted to spend one extra second doing something inconvenient.

So the work that cannot be copied is making someone feel a flicker of delight inside an experience that is, by its nature, a chore. That is taste. That is judgment. That is knowing what good looks like and why, and being able to take a thousand scraps of inspiration and shape them into something that serves one real human being. A model gives everyone the good enough version. A designer is the one who refuses to stop there.

Here is the part I love. Design was never even a workstream on the Task Force. I did not name it. I did not assign anyone to it. They made the shift on their own anyway. They handed deck-making over to the org so people could serve themselves. They stopped standing guard at the asset factory. And they took every hour they got back and poured it straight into the craft.

That is the whole move. Give away the thing that does not need you. Reclaim the time to do the thing only you can do. Let it go. It was never serving you.

Go Be Unrepeatable

So if you are a designer and you are quietly spiraling right now, hear me. I mean it.

Your job was never to make the flyer. It was never to push pixels on command for whoever asked loudest. Let that go. It does not serve us, and honestly it never fully served you either.

You can sit shoulder to shoulder with a developer and a product manager and build an entire experience, beginning to end. Not the version you slam out against a deadline, stuffing in whatever the timeline barely allows and calling it shipped. The version with room to breathe. The version a frontier model could not dream up in a thousand tries, because dreaming it up requires actually understanding the person on the other side of the screen.

You can watch how people really behave inside your work. You can know the human you are building for, which is the one thing a model will never care about the way you do.

We are at ASU to make education reachable for everyone, no matter where they are starting from, in the most human way we know how. The designer who can build a whole experience around that, and not just decorate it on the way out, is worth more right now than they have ever been.

The exact thing that scared you, that anyone can now make something that looks like your work, is the same thing that makes you essential. Looking like the work was never the work.

So learn what to set down. Then go five steps past where anyone can follow.

Go be unrepeatable.

I am the leader. I asked for this. I built the Task Force. I named the workstreams. I declared the whole thing out loud. So of course I think it is worth it. You should be a little suspicious of how convincing I sound.

The people you actually need to hear from are the ones we asked to come with us. The ones who did not get to choose. The ones quietly rethinking careers they spent years building, because we told them the ground was moving and asked them to trust us.

They know things about this that I will never see from where I sit.

So I am handing them the pen.

First up is Stephanie Morse, Assistant Director of AI Content Strategy and Systems, who leads our content team across product, programs and user experience.

I keep describing the edge.
She is the one walking it.

-Gulley

By Stephanie Morse

A few months ago, I got everything I’d been asking for at work: AI tools, blocked time to use them and permission to figure out what was possible. What I didn’t expect was how quickly that freedom would become its own kind of pressure.

I changed my Slack status to “AI experimenting - Slow to respond,” blocked my calendar and spent days loading my work life into Claude, trying out new prompts and experimenting with Cowork.

The wins came quickly. A detailed brand guide built in minutes. Agents optimized our knowledge base while we focused on strategy. Program pages scaled faster than we’d ever managed during our busiest time of year. A content system of skills that could generate nearly any asset we regularly needed to produce. Projects that had been in our backlog for months, or even years, finally felt within reach.

The part that everyone talks about was real. But it was also just the beginning of a harder story.

What I didn’t understand yet was that crossing into new territory and knowing what to do there are completely different problems. A new frontier was ahead of us. Now what? Nobody gave us a map because nobody had one.

The Part After The Part Everyone Talks About

For years, the constraints on what my team could produce were real and familiar: Bandwidth, headcount, time, the sheer physical limit of how much a small team of humans could think, write and revise in one week. We were good at working within those constraints. We’d built our whole operation around them.

Those constraints were also clarifying and comfortable. You knew what to prioritize, what to expect and what the team was capable of accomplishing each and every week.

But now those constraints have completely changed. And changed in ways we don’t fully understand yet. The comfort faded away, and we were left with hard questions: What do we want to focus on? What do we actually want to build? What does quality mean now? What is our job?

There’s also real pressure. There’s external pressure to show what AI is unlocking, the efficiency we’re gaining and to always be keeping up. And there’s an internal pressure that’s even more surprising and harder to name: When your team can do more than ever before, you feel the weight of every choice. The accountability shifts, and every “no” feels heavier than it ever did before.

Even team meetings and one-on-ones changed. Instead of focusing primarily on projects, timelines or priorities, something new started to emerge. Conversations about identity. About what it means to be a writer and content strategist when you can generate content at the press of a button. About when AI is genuinely saving us time and when, honestly, it isn’t yet. About the role of our team in this new reality.

These questions don’t have easy answers. And they don’t go away just because you’re excited about the possibilities. But they deserved answers, so we started working on mapping this new frontier.

The Tools Were the Easy Part

We knew we’d have a lot to learn. But what surprised me was how much of what we already knew still mattered — just differently than before.

From doing to directing: As writers, we’re used to controlling the words. We have opinions about every sentence. We agonize over word choice and analyze every comma. That’s not a bad trait in a content strategist. For most of my career, it was essential and what I did every day.

However, AI asks you to give up that control in a meaningful way. You build an AI system that creates content, and suddenly you’re no longer writing every word or thinking about every sentence.

All of that hard-won knowledge about brand voice, grammar, strategy and what makes content actually work doesn’t become irrelevant. It becomes the foundation on which you build the AI system. The expertise moves upstream, from the sentence to the brief you feed AI and from the draft to the guardrails you build into the AI system.

We still have humans in the loop reviewing AI-generated content. But we’re no longer writing every sentence. And for a team of people whose professional identity is built around caring deeply about every word, that’s a more personal shift than it might sound from the outside. You have to learn to trust the system you built. Which means trusting that your expertise transferred — that it’s still there, just upstream now.

From creating to building:

We used to start nearly every project with creative ideation. We went from a blank page to brainstorming, outlining, drafting and revising. Now more often than not, we start by building. What systems do we need? What agents or skills can we create to complete this at scale?

That’s a hard change for a team of writers because not only is the workflow completely different, but nothing in a writing career explicitly trains you to think this way. Writing develops taste, voice and judgment.

The good news is that the editorial instinct is exactly what distinguishes a content system that works from one that’s merely functional. You’re not starting over. You’re learning to apply what you know in a different way. But that shift has to be learned. And it’s a genuinely new skill for people who became professionals by mastering a different one.

From writing for humans to machines:

For most of my career, content had one audience, and that was other humans. Every decision about what content to include, what to leave out, how to organize a page and where to place a CTA was made to create the best possible user experience. Yes, there was SEO and search algorithms, but we were still writing for other humans and not the algorithm.

Now, large portions of our content aren’t meant to be read by students at all. They’re written for the AI tools that serve information to students and support staff — systems that decide what a reader finds before the reader ever gets there.

Structure, consistency and clarity used to be about the reader. Now they’re also about the machine deciding whether your content gets surfaced accurately (if at all). Every word is doing double duty. The craft is still there. But content is infrastructure now. Content is data. And we’re learning to write for both.

Everyone Else Is Figuring It Out, Too

Just when you start to find your footing, you realize everyone around you is figuring it out, too. You’re not just managing your own transformation. You’re navigating everyone else’s at the same time.

My team received AI-driven contributions and feedback. We’ve been part of projects where we intentionally led with human-driven content while other collaborators led their part with AI. We had to figure out how new workflows fit together in real time, with no playbook.

This creates real friction and new overhead that didn’t exist before. Workflows and processes that had been refined over months and years no longer work quite the same way. Problems come up at unexpected times. Scaling isn’t always as easy as it looks from the outside. You go into each project not quite sure what to expect. Managing all of that with grace takes a lot of energy.

But the other side of that is real, too.

There’s something genuinely energizing about experimenting alongside other teams who are figuring it out at the same time you are. And when things come together just right, you can accomplish things that wouldn’t have been possible before.

Recently, we launched a tool to help students navigate a significant change in federal loan policy. With AI prototyping, we went from idea to launch in weeks, not months, at a moment when students needed clear information fast.

That’s what it feels like when the new ways of working click.

There have been plenty of bumps, frustrations and wrong turns along the way, and there will be more. But that project is one reason why we keep going.

The Best Thing We Did

What the highlight moments don’t show is the fatigue. Learning new tools, redesigning workflows, rethinking your professional identity — all while hitting your normal project deadlines — accumulates faster than anyone warns you about. I feel it. My team feels it. And I don’t always handle it well. Sometimes I add to the pressure when I should be absorbing it.

But one thing we got right as a leadership team was giving our team their afternoons back for AI upskilling. We called it Recess.

Some of what came out of those sessions was immediately useful. People built new AI tools and developed skills they’re still using every day. A big part of it was just figuring things out side by side.

We also had real conversations about what quality looks like when you’re working at scale, when AI is genuinely the right tool and when it honestly isn’t and what the human judgment layer is actually for. Those aren’t questions with clean answers. But having them together, explicitly and honestly, meant that when we started making real decisions, we had a shared language.

Asking a team to change how they work is significant. Asking a team of content creators to reconsider their relationship with writing is deeply personal. The least we could do was make space for that transition to feel human and to signal that learning together was not a distraction from the work. It was the work.

To My Team, and Maybe Yours

Here is the most honest thing I can tell you about leading through this: I don’t have a complete map. I have a direction (sometimes), a set of values and a team I trust. Most days, that feels like enough. Some days it doesn’t.

I’ve asked my team to come along with me on something I can’t fully describe yet. I’ve asked them to learn skills they didn’t sign up to learn, to rethink work they spent careers refining, to stay curious in the face of real uncertainty. I’ve tried to be honest with them about what I know and what I don’t. I’ve tried to make space for the hard questions without pretending I have all the answers.

What I know is this: Our team can do more and lead more than we ever could before. That is genuinely true. It didn’t come free. There were uncomfortable conversations, identity shifts and moments of real not-knowing. There are still more of those ahead.

But we have a new frontier. And we’re still here, scribbling out a map together. We don’t know exactly where we’re going. We’re showing up anyway. Turns out that’s the whole job.

Taking a break from my recent AI task force group overviews, to share out something happening in real time. I had another post I wanted to share, but this all happened within 24 hours, and so here I am, writing about it. Here it goes.

I used to think the danger was falling behind. Now I lie awake worried we are getting ahead of ourselves and calling it progress.

Last night I reviewed a product my team built faster than we could have a year ago. Faster than I thought possible. They used AI exactly the way I have been begging everyone to use it. By every metric I preach, they did everything right.

And it was full of holes.

It Looked Done. That Was the Problem.

The work looked finished.

It rendered. It was polished. It sat there on the screen looking like a real product, and looking like a real product is a feeling. You feel like you made something bigger than you could have made alone. And you did. You just made a more incomplete version of it, faster.

Then I started asking the questions we used to ask before anything got built.

How does an administrator manage members. How do we see token consumption by user. What are the required fields, what data does a user actually see when they deploy their solution, what happens to every input that matters and every one that doesn’t. How does a person even create an account, and how many ways will they try. How do we test this. What does success look like if everything goes right.

None of it was there. Not wrong. Not there.

These are not edge cases. This is the job. This is the unglamorous, deeply human part of building a product, the part where you sit with how a real person will actually use the thing in all the messy ways they will use it, and you make a hundred small decisions nobody will ever thank you for. That used to be where we spent most of our time. Because that was the part that mattered.

The Work Wasn’t Offloaded. It Disappeared.

I keep hearing that tired people hand the boring parts to AI. That is half of it. Change fatigue is real and I am watching it settle into my teams like weather. When you are tired and there is a tool right there that will take the parts you do not enjoy, you hand them over.

But that is not what happened here. Nobody decided the admin view did not matter. Nobody weighed it and skipped it. It simply never came up. The tool produced something that looked complete, and because it looked complete, completeness never got questioned.

That is the part that should scare us. The failure was not bad work. The failure was plausible work that nobody checked. AI does not know there is an administrator who needs to manage members unless someone tells it. It does not know the seven ways a user will try to create an account unless someone has sat in that chair and felt the friction. It will confidently build you something beautiful and incomplete, and it will never once flag the absence. Absence is invisible to a thing that only knows what it was given.

So the human in the loop steps out of the loop without noticing. And the cost does not vanish. It moves downstream, where it is more expensive to fix. Less time now. More work later. To produce the quality we used to get the first time.

Can AI code? Obviously. Can it hold every complexity of a product without context, without a human in the loop, without someone who actually owns the experience? No. And the moment it cannot, speed stops being an advantage. Speed becomes a loan. And the bill always comes due.

The Ground Is Moving While We Run

This week Anthropic published a piece called “When AI builds itself.” The first thing I saw was the graphic. Chat, then agents, then workers, then a future where the AI improves itself. The length of tasks AI can reliably complete on its own has been doubling roughly every four months. The thing builds the thing now. anthropic

I sat with that next to my broken PRD and felt the floor tilt. Because here are two true things at once. We are not yet using AI to anything close to its full advantage. And the definition of full advantage already changed while we were learning the last one.

The minute you feel good at something, something underneath you has already moved. The minute we get comfortable and consistent, we are behind. There is no version of this where we slow down to feel safe. The race does not stop because we are tired.

So We Run Smarter

If we could slow down, it would be better for all of us. Time to breathe. Time to learn what good looks like before we are asked to ship it. That is not coming. So the question is not how to stop. It is how to keep borrowing without going bankrupt.

Here is the only answer I have found. You cannot run every leg of the race yourself. The reason that PRD had holes is that one person tried to be the engineer and the designer and the product manager all at once, and let the tool quietly fill every seam. The seams are exactly where it falls apart. The login nobody mapped. The permissions nobody scoped. The test nobody wrote.

When one person absorbs the whole team, they finally understand what each person was carrying. The specification. The detail. The hundred small acts of care that added up to something a real human could actually use. You do not see it until it is gone.

So we pass the baton. We hand off the judgment when it is not our leg to run, so we have fresh judgment when it is. That is what a team is for. That is the thing AI does not replace, it reveals.

We are failing fast, and I would not trade it. I have learned more in three months of getting it wrong than a year of getting it right would have taught me. But failing fast is not the lesson. The lesson is what it costs, and who pays.

Speed is a loan. We are going to keep borrowing, because standing still costs more. The only thing that decides whether we can afford it is whether a human is still awake at the moment the machine says it’s done.

So what does that free us up to do?

That is the question underneath everything we built with our project management team. And the answer changed how I think about transformation entirely.

But getting there required something most people skip over. Curiosity. A willingness to take apart how you work, look at every piece, and ask honestly what still makes sense. To understand where your time actually goes. To put down the wall that says AI is coming for your job and pick up the much more interesting question: what becomes possible when it handles the parts of your job that were never really yours to begin with?

And that is why we picked project managers as our second 30-day group.

Why PMs?

Project managers are the connective tissue of any organization. They are in every room, on every call, holding the threads that keep complex work from unraveling. They know the deadlines. They know the blockers. They know who needs to talk to whom before a decision can actually get made.

They also spend an enormous amount of their time doing work that has nothing to do with any of that.

Writing down every spoken word in a meeting. Transcribing requirements. Chasing context they were not sure they fully captured. Parsing technical nuances from conversations they were not trained to understand. Documenting, organizing, re-documenting. The administrative weight on a project manager is real. And it was quietly burying the strategic capacity we actually needed from them.

We knew that if we could lift that burden, we would not just help the PMs. We would feel it everywhere. Because when the connective tissue of an organization gets stronger, the whole body moves better.

That is why we picked them second. Not because they were the easiest group. Because they were the highest leverage one.

Focus

Before I could put the right PMs on this work, I had to make some hard calls.

Part of my role in standing up the Task Force was going deep into what people were actually working on across every group I was considering. Not surface level. Deep. I needed to understand what was essential, what was legacy, and what was sitting on someone’s plate simply because it had always been there.

What I found in the PMO was two project managers mid-project. Work in progress. Timelines in motion. Teams depending on them.

Those PMs had to stop work they had invested in, set down something unfinished, and redirect entirely. That kind of pivot carries a psychological weight. It feels like abandonment. It can feel like failure, even when it is not.

Every resource decision in a transformation like this is a prioritization decision. And prioritization means someone does not get what they expected. That includes teams I left temporarily without PM support. I made that call knowing the gap it would create. Because the alternative was under-resourcing the work that had the highest potential to scale.

You cannot lead a transformation from a place of discomfort avoidance. At some point you have to know the work well enough to make the hard call. And then make it.

Where is the map?

Project managers are wired for clarity. Give them a problem and they will build you a structure. Give them a deadline and they will reverse engineer every step to meet it. But this was different. We were not implementing a known system. We were inventing one. The tools were new. The integrations were untested. The templates did not exist because nobody had built them for this context yet. Every week revealed something we had not anticipated and required a decision we had not planned for.

The discomfort of not having a playbook is not a problem to solve. It is the condition of doing work that has never been done before. If the playbook already existed, we would not need a Task Force. We would just need a project plan.

Finding peace with uncertainty is not a soft skill. In a world moving this fast, it is the most important operational capability a team can develop. We are writing the playbook in real time. That is the job. And the teams that embrace that, instead of waiting for someone to hand them the answer, are the ones who end up writing the playbook everyone else will follow.

Supercharge

We brought in tools to change the meeting experience entirely. Not just to take notes, but to free our PMs to actually be present. To listen. To think. To lead the room instead of transcribe it. AI note-taking tools meant they could walk into a meeting prepared, knowing what was discussed last time, what was left open, what needed to move.

But notes were never the finish line.

We needed templates. We needed those templates connected to Jira, so action items became tickets on the right boards assigned to the right people automatically. We needed Slack integrations so updates flowed without anyone having to remember to send them. We needed Slack canvases and meeting note folders as living knowledge bases on every project, so we could query these databases to receive real answers in real time, through AI chatbots.

We needed the whole system. Not just the shiny tool sitting on top of it.

It required someone willing to go all the way in. Aly Rassel was that person. She leaned into the emerging technology side of this with everything she had and quietly became an expert across tools. She did not wait to be asked. She just went. She was a leader in the space, defining what it meant to be a project manager in this new AI-enabled world.

The PMO is not what it was two months ago.

Because of this, our project managers can once again be strategic operators. They are the layer above the work, pulling threads across all five Task Force groups, surfacing what matters, connecting dots that would otherwise stay disconnected. We are using these tools to scrub through Jira backlogs, organize tickets by label and component, understand who requested what and why, and make smarter prioritization decisions at both the PMO and operational level.

The visibility we have now into what we are doing, why we are doing it, and whether we should keep doing it is unlike anything I have seen.

And that is the point. The Task Force was never just about the five workstreams. It was about proving that when you give the right people the right tools and the room to work differently, you get something that cannot be undone. A new floor. A new baseline. A new standard for what is possible.

Today, I am attending our first AI Day with my full leadership team. We are dedicating an entire day to doing the work, learning the tools, and seeing what emerges when you stop talking about transformation and actually do it.

It started two months ago with a conversation and a new focus.

It started with permission to be bold.

Some things are worth repeating.

Still Loading is a real-time leadership diary by Amanda Gulley, Chief Product and Experience Officer and Deputy of AI and Automation at EdPlus, Arizona State University.

Everybody wants to be transformed. Nobody wants to be the team that gets chosen to do it before anyone else has proven it works. Before there is a playbook. Before there is a success story to point to.

We needed someone to go first anyway.

Six months ago I sat across from a content team that had already made up their mind about AI. Not that they wouldn’t use it, because they were using it daily, but that they believed only their team could write quality content. They had watched what happened when people stopped caring about quality and just started generating. They had seen the slop. And they had decided, quietly and collectively, that they were not going to be part of that story.

I needed them to change their mind.

The Climb

When we launched the task force, we knew the first sprint would set the tone for everything that came after. Not just operationally. Psychologically. The first team we chose would either prove that this way of working was possible or give everyone else in the organization a reason to wait and see.

That is a lot of weight to put on one group of people.

We needed a workstream where the need was undeniable, where the technology was genuinely ready, and where the right leader was in place to carry the team through something hard. We needed a win. But we also needed it to be real. A manufactured success story wasn’t going to move anyone.

Content kept coming back to the top of the list.

We didn’t have enough writers. We never did. The volume of content EdPlus needs to operate at scale, across programs, platforms, emails, student touchpoints, knowledge bases, is not a hiring problem. It never was. AI was always going to be part of this answer.

But the content team had already seen the slop.

They knew what AI-generated content looked like when nobody cared. When someone just hit generate and shipped it. They had watched it flood the internet. They had watched it show up in places it had no business being. And they were not interested in being part of that story.

That wasn’t resistance. That was professional standards. That was people who cared about their craft saying: not like that.

I respected it completely. And I also knew we had to move. Which meant before we could ask anything of this team, we had to answer a harder question first. Not “can AI write content?” It already can. The real question was:

What is content?

We keep hearing the same thing in every AI conversation. You can’t build an AI strategy without clean data. Get your infrastructure right first. Get your data house in order.

I believe in data infrastructure. And I also know we have launched several successful pilots without having all our ducks in a row. So that can’t be the whole story.

Here is the part everyone forgets. In a world of generative AI, content is data.

Your content is not just what you publish. It is what your models learn from. It is what surfaces when someone asks an LLM a question about your institution. It is the personality, the tone, the knowledge base that provides real service because it knows how to answer a student’s most pressing question at 11pm. Content is becoming one of the most important data sets you own. And most organizations are not treating it that way.

You do not have to wait for perfect infrastructure to start. You have to start treating your content like it matters as much as your data. Because it does.

Once we framed it that way, the ask changed completely. We weren’t going to the content team to say “AI can do your job faster.” We were going to them to say “your expertise is more valuable than it has ever been, and here is the entirely different way we need you to use it.”

Something Borrowed, Something New

Everything they had spent years learning still mattered. The way they thought about search engine optimization, about what makes content findable and trustworthy and worth a reader’s time. The instinct for when a sentence earns its place and when it doesn’t. The deep knowledge of what EdPlus and ASU sound like, what our students need to hear, how we show up in the world. None of that was going away.

We just needed it to go somewhere new.

Instead of editing for clarity, you analyze for gaps. Instead of asking “is this well written,” you start to ask “is this accurate, is this consistent, and what does this output tell us about what our model doesn’t understand yet about who we are.” Instead of manually building every program page field by field in a CMS, you figure out how to automate it. Instead of being the team that produces content, becoming the team that sets the standard for how content works across the entire organization.

Content “skills” built inside our AI environment. Tested, broken, improved, and eventually handed to everyone else as the bar.

That is a completely different job. It looks nothing like what they were hired to do. And we were standing in front of them asking them to trust that the version of themselves on the other side of it was worth the crossing.

Which meant everything that came next depended almost entirely on one thing: the person standing at the front of the room when we made that ask.

“Trust Me”

Allison leads our content team. And before any of this started, I asked her to trust me.

Not with a guarantee of what it would look like on the other side. I asked her to trust that the direction was right, that the team would be better for it, and that I would not ask them to go somewhere I didn’t believe in completely. That is a lot to ask of someone. Especially when you are asking them to carry that trust into a room full of people who are looking to them for certainty you haven’t given them yet.

She didn’t package the uncertainty into something more palatable or wait until she had better answers before she started moving. She brought her team the honest version. This is hard. I don’t have it all figured out. We are doing it anyway. And then she showed up the same way on day 14 as she did on day one, even when the doubt was louder than the momentum.

That is a kind of courage that doesn’t often get shared. It’s the quieter, harder courage of staying steady for your people when steadiness is the last thing you actually feel.

And they met her there. Every single one of them.

Reinvention

They have at minimum 3X’d their capacity. Not because we pushed them harder, but because they fundamentally changed how they work and what they work on. The manual inputs, the repetitive production, the work that required their hands but not their minds, they are building the systems that handle that now. They are becoming data stewards. Getting more technical every week, not because we required it, but because they followed the momentum and it led them somewhere they hadn’t expected to go. I watch them now, and I see as much of a technology team as I see a content team. That boundary dissolved somewhere in the middle of the sprint, and nobody made an announcement about it. It just happened.

We haven’t even opened the tools to the broader organization yet. When we do, they will be the architects of how quality scales across EdPlus. The ones who went first built the skills, set the standards, and figured out what everyone else would eventually learn. They came out the other side changed. Watching it happen changed something in me too.

Wayfinding

The thing about going first is that nobody gives you a map. You are not following someone else’s courage. You are making your own, in real time, with people watching, on days when you are not sure it is going to work.

And I have learned that you cannot manufacture that kind of moment. You cannot engineer it into a sprint plan or a change management framework. What you can do is create the conditions for it. You can choose the right team. You can ask the right leader. You can say “trust me” and mean it and show up every day like you meant it.

That is the thing about transformation that nobody tells you. If you do it right, it will exceed what you planned for. It will ask more of people than you thought was fair to ask. And they will rise to it in ways that change not just the work but who they understand themselves to be.

If you are leading transformation right now and you are wondering whether it is worth the weight of it, whether the late nights and the hard conversations and the grief of letting go of what was are adding up to something real, I want you to know this.

It is. But only if someone is willing to go first.

Find your Allison. Be someone’s Allison. And then get out of the way and watch what people are capable of when they finally believe the ask is real.

An artifact we created during the process to align on where we were going.

Swim lanes. Pillars. A roadmap with color-coded quarters. A slide that says “AI-Enabled Workforce” in a bold font next to a stock photo of people collaborating around a laptop. Everyone nods. Someone says “this is really exciting.” The meeting ends.

And then nothing changes.

I have been in those rooms. I have built those decks. And I say this with full accountability: we were naive. Not careless, not incompetent. Naive. We thought we could predict what this technology would disrupt. We thought we had enough runway to be methodical. We thought six months of planning would give us a foundation sturdy enough to build on.

What we didn’t account for was that the technology would stop waiting for us.

We Are No Longer Thinking in Quarters

A year ago, a major model update was a moment. You planned for it. You adjusted. You moved on. Now a model can evolve multiple times in a single week. What was cutting edge on Monday is table stakes by Friday. The pace is not accelerating. It has already accelerated. We are inside the acceleration now, whether we planned for it or not.

That changes everything about how you lead change.

Traditional change management is incremental by design. You assess. You pilot. You scale. You measure. It assumes you have time to be right before you move. AI transformation does not give you that luxury. You are not steering a ship. You are trying to catch a current that is already moving faster than you can paddle.

We were not ready to ask that question six months ago. We are asking it now.

The Bigger Thing Nobody Is Saying Out Loud

Here is what keeps me up at night. And I don’t mean that metaphorically.

I heard a NPR and CNN segment today about declining reading and math scores in children. The instinct is to blame COVID. COVID was real, COVID was devastating to learning. But the data shows the decline started before the pandemic. It started with the rise of social media, short-form content, and the amount of time children spend on screens that are designed, literally engineered, to overstimulate their developing brains.

Kids' test scores began declining way before COVID. These schools are making gains

My own kids will not have phones for as long as I can possibly hold that line.

But here is the thing. I lead an organization whose entire purpose is student success. We are trying to help people complete college degrees and build careers and lives that matter. And we are doing that in a world where the foundational skills of learning, reading, sustained focus, are being quietly eroded by the same technological forces we are trying to harness for good.

We needed to think exponentially. Not incrementally. Not “how do we add AI to what we already do.” But “how do we reimagine what we do, at the root, for a world that is changing faster than any of us can fully see.”

That is a different conversation. That is a harder conversation. And it is the only one worth having.

So We Named It

When the AI Task Force launched six months ago, the first decision that mattered was not the technology. It was not the workstreams or the budget (no new buget given) or the timeline. It was the name.

1. Name It. I have already publicly admitted that “Task Force” is my biggest regret. We should have called it something cooler. But naming it at all? That was the right call. Because naming something makes it real. It signals to your organization that this is not a side project. This is not the thing we will get to after the roadmap settles. This is the work. And naming it was only the first step.

2. Date It. Six months. That was the commitment. Not “we are going to explore AI transformation over the next year or two.” Six months, five workstreams, real people, real accountability. A date forces honesty. It forces prioritization. It forces you to stop optimizing the plan and start doing the thing.

3. Mean It. Top down, without ambiguity. Leadership said: this is happening, this is why, here is what we are doing. But here is the part that actually made it work: the strategy is bottom up.

We did not hand down a playbook. We pulled people up. We went into teams and asked: what are your pain points? What breaks every single week? What would you change if you could? And we brought in the AI evangelists on the technology teams to show what was possible even outside the furthest boundaries of the team’s suggestions. We found the people in those teams, sometimes the most junior people in the room, who were not yet shaped by “this is how we do it here.” The ones who were still asking “but why, though?” We made them the fire starters.

Because that is how you spread this. You do not broadcast AI transformation from a stage. You ignite it in small rooms, with the right people, and you let it burn outward.

Everyone Is a Technology Team Now

I read something once that said technology teams would eventually become more like IT, the department that makes sure everyone else has what they need to do their jobs. I used to resist that framing. Now I think it might be exactly right, and not as a demotion.

Our job is no longer to be the only people in the building who know how to use the tools. Our job is to make sure everyone in the building can. Every function. Every team. Every person. AI capability is not a competitive advantage if it lives in one corner of the org chart. It is a liability.

You are watching companies do massive layoffs right now, cutting deeply, and then turning around and hiring back at significantly higher price because they want AI-enabled talent. I understand the math they think they are doing. I do not accept the premise. There is no reason, none, why you cannot develop that capability in the people you already have. The people who know your mission, your customer, your culture. The ones who have already earned trust.

It takes time. It takes intention. But it is not impossible. It is a choice.

A Name That Deserves to Be Said

None of what I just described would be moving at the speed it is without my colleague Auryan.

Auryan has been a disruptor on my team for years. The kind of person who does not think inside the walls of how things have always been done, because he genuinely cannot. His brain does not work that way. And for a long time, in a lot of organizations, that quality gets managed down. People get told to work within the system, to be realistic, to stop asking questions that make everyone uncomfortable.

We did the opposite. We pointed him at the hardest problems and got out of the way.

The result is that his thinking is now spreading. The fire is catching. Ideas that lived in one person’s head six months ago are now showing up in how multiple teams work, how they approach problems, how they see what is possible.

Hire people who do not fit neatly inside your existing walls. Protect them when the culture tries to smooth out their edges. And then watch what happens when you finally give them a real problem to solve.

The Advice, As Simply As I Can Give It

We are not done. We are not even close to done. But we have learned enough to say this clearly:

Name it. A real name. A public commitment. Something your organization can point to and say, that is the thing we are doing.

Date it. Not a soft target. A real one. Dates create urgency. Urgency creates movement. Movement creates learning you cannot get any other way.

Dedicate people to it. Not as a side project. Not as the thing they do between their real responsibilities. Actual people with actual time and actual permission to fail forward.

And mean it. Relentlessly. Top down so people know it is real. Bottom up so people know it is theirs.

The technology will keep moving whether you are ready or not. The question is whether your people will be moving with it, or watching it from the sidelines, waiting for someone to tell them what to do next.

We decided we were not going to wait.

I waited to be told it was okay to take the bold swing. To push the idea. To move before everything was perfectly in place. I kept looking for someone to give me permission to do the thing I already knew needed to be done.

That permission rarely came. And when it did, it was usually too late.

I think about that a lot now. Because the thing I’ve learned -- the hard, slow, humbling way -- is that nobody is going to give you permission to lead boldly. You have to decide to do it anyway.

It’s where the task force really begins.

Put Me in Coach

We began planning our annual budget planning in January of this year.

You make the list. You build the case. You sit in the room and advocate for everything you need. And then you wait.

I’ve done it every year. This January, I couldn’t do it anymore.

EdPlus built its identity on speed. On experimentation. On being the place inside a major research university where the rules bent a little, where you could try something and fail and try again. We were the startup inside the institution.

But startups don’t stay startups. We had grown to eight hundred people. The systems were complex. The integrations were layered. And everywhere I turned, someone was explaining why something couldn’t happen yet.

I was watching things get built in days that were taking us months. And the word I kept hearing was “can’t.”

I was tired. Tired of hearing that people didn’t have time.

Because we always have time for what we choose. That’s the truth that’s uncomfortable to say out loud in a professional setting, but it’s true. We make choices. Every week, every quarter, every year, we decide what we prioritize. And for too long, we had been choosing the urgent over the important.

I also knew something that wasn’t talked about enough: we were building technology for students, but our staff was running on duct tape and willpower. Our teams weren’t being resourced with the same intentionality as our products. And if we were going to ask people to work differently, we had to show up differently for them first.

I shared my realization with my boss, and said “put me in coach”. It was time to make a bold decision, and as a leader, I knew the road ahead would be the toughest climb to Everest. But, it is what we needed, and the only way forward.

The Task Force was Born

My new title -- Deputy of AI and Automation -- wasn’t a new role, it was not a promotion. It was a declaration to our organization. This is real. This is resourced. This is not optional.

We announced it to our leadership team before we had all the answers. I want to be clear about that. We named it, we stood in front of our team, and we committed to something we were still figuring out. Because waiting for certainty in a moment like this isn’t caution. It’s a choice. And I was done making that choice.

I selected five workstreams for different reasons. Three of them -- Coaching, Instructional Design, and Product Engineering -- were our biggest long-term bets. The places where we believed AI could most fundamentally change how we do our most important work. And the work that was necessary to achieve our most impactful initiatives toward our 2030 goals.

But we also needed to prove something. Fast.

Because the resistance I was hearing wasn’t that people didn’t believe in AI. It was that people were already using it -- and they weren’t wrong. The problem was how they were using it. As an assistant. A shortcut. A faster way to do the same things.

We needed people to start using it like a team member.

So we chose two other 30-day sprint groups -- content and project management -- specifically designed to show what was possible when you actually changed the way you work. Not just the tools. The work itself.

Spoiler alert: We are currently 2 months in, and it’s quite magical what happened next. Stay tuned for the next article.

The Part Nobody Talks About

So we built the plane while we flew it. And I think that was the right call, even when it was uncomfortable.

What I didn’t anticipate was how much my leaders needed each other.

The harder I pushed, the more I noticed something fraying. The conversations my leaders were having separately, in their own silos, about their own teams’ fear and frustration -- those conversations weren’t making it into the room where we were all together. And the glue that held our leadership team together was starting to come undone.

The inflection point wasn’t a strategy session. It was when I finally brought everyone together and we stopped performing alignment and started being honest. We questioned each other. We said the hard things. We held each other accountable in the open, not just in the quiet one-on-ones where it’s easier to be kind than to be real.

That is when the culture started to shift.

We are still actively breaking down the silos, the swimlanes, the handoffs. As I am typing this, these conversations are happening right now in real time. But then I remember, I was once a designer, who became a UX researcher, who became a product manager, who then became a leader over many other disciplines. Why should we hold back any of our teams from being more? It’s not about needing more; it’s about accepting that our teams have more to give, beyond their lines and boundaries, and it’s time we give them the space to create and innovate and to become more.

What the Task Force Actually Is

Looking back, the task force wasn’t about a set of workstreams.

It was permission.

Permission for our teams to stop waiting for everything to be figured out before they tried something new. Permission to experiment, to fail fast, to believe that their boldest ideas were worth pursuing right now. Permission to be the kind of organization we’ve always said we are.

I spent years early in my career waiting for someone to give me that permission. I know now that nobody was going to hand it to me. I had to claim it.

So that’s what we did. For ourselves. And for every person on our team who needed someone to say -- yes, now, go.

The most important work is ahead of us. We stopped waiting to start it.

My only regret? We should have given it a cooler name.

A note on the visuals: some of what you’ll see throughout this series are real slides we used during our communications to our teams. No polish added. This is what it actually looked like.

Next issue: how we built the task force -- the groups, the structure, and the decisions that shaped everything that came after.

None of it compares to what we’re going through right now.

I don’t say that to be dramatic. I say it because I think a lot of leaders are privately feeling the same thing and not saying it out loud. And I’ve made a choice, maybe an unconventional one, to say it out loud anyway.

We went to a product summit in 2024 hosted by where openly shared this presentation on how the talent stack is collapsing. I didn’t have the foresight then to fully grasp this reality. But I do now.

The Feelings Of Transformation

Over the last few months, LinkedIn has become something of a diary for me. Not a highlight reel. Not a place to share wins and polished takeaways. A place to process, in real time, what it actually feels like to lead a team through AI transformation.

The raw version. The one where I’m not sure I’m getting it right. The one where I’m asking my team to let go of things they’ve spent years building, and feeling the weight of that ask every single day.

I’ve found it cathartic. More than that, I’ve found it useful. Because when I share what we’re going through, I hear from other leaders who are in the exact same place, or where we disagree (which I appreciate the most). Quietly overwhelmed. Excited and terrified at the same time. Trying to figure out how to move fast without losing the people they care about.

Everything We Built Is Changing

Here is what I know to be true right now: everything we built is changing faster than we can keep up with.

The domain expertise that defined our teams. The workflows we optimized over years. The skills we hired for, trained for, and promoted for. All of it is being renegotiated. Not because it wasn’t valuable. But because the world shifted underneath us and it is still shifting.

That is an incredibly hard thing to ask people to accept. It’s hard for me to accept, and I’m someone who genuinely believes in where this is going.

The discomfort is real. The grief over what’s changing is real. And I think we do people a disservice when we skip past that and go straight to the excitement.

Why I Am Turning This Into a Newsletter

So we’re taking it a step further.

We’re documenting the process. The thinking, the decisions, the things that worked, the things that didn’t. Not as a case study after the fact, but in real time, as we’re living it.

Follow our process or don’t. Borrow what’s useful and leave what isn’t. But I hope it helps someone.

Especially our students. The ones we serve every day who are trying to figure out what careers look like in a world that none of us have fully mapped yet. They deserve to see leaders being honest about the uncertainty, not performing confidence we don’t always feel.

We are still preparing for this world. We are building the plane while flying it.

We might as well document the flight.

Subscribe now