Their first reaction wasn’t “How do I buy more?”

It was “How do I turn some of this off?”

You build something valuable enough that customers use it heavily, and the monetization model turns your most enthusiastic adopters into your most anxious cost managers.

The obvious conclusion is that credits are bad. That misses the more interesting problem.

AI pricing is becoming a trust problem before it becomes a monetization problem. And almost every enterprise software company building serious AI will have to confront it.

The vendor isn’t wrong either

AI changes the economics of software. For most of SaaS history, the marginal cost of another click or another workflow was small enough to ignore. We built an entire industry around encouraging customers to use more: activation, adoption, DAU, MAU, feature penetration. Higher usage was almost always good news.

Agentic AI breaks that equation. An agent continuously reading thousands of calls, researching hundreds of accounts, and reasoning across large datasets creates real variable cost, and that cost grows with usage.

So vendors face a perfectly rational problem. Raise the subscription price for everyone, including customers who barely touch the expensive capabilities? Absorb the inference costs and hope model economics improve? Or charge by consumption?

Nothing about choosing consumption is unreasonable. The problem begins when a pricing mechanism that makes perfect sense inside the vendor’s spreadsheet creates exactly the wrong behaviour inside the customer.

A good monetization model should make customers want to consume more value, not ration it.

AI has broken an old psychological contract in SaaS

For twenty years, SaaS taught customers something simple: I buy access, and you want me to use it.

The customer paid for the seat. The vendor then worked relentlessly to increase adoption: Customer Success encouraged usage, Product removed friction, executives celebrated engagement charts. The incentives were beautifully aligned.

AI introduces a strange new message: Please adopt aggressively, but not too aggressively.

Yesterday, Customer Success was pushing teams to embed AI across their workflows. Today, the customer discovers that every one of those workflows draws down a finite pool of credits. And this lands hardest when customers first experienced the AI as part of the product, built habits around it, and only later encountered a separate economic model for the usage they’d already normalized. From the vendor’s perspective, this is packaging evolution. From the customer’s perspective, the rules changed after they started playing.

That is why an AI pricing change can never be a rate-card exercise.

Pricing changes are trust events.

The moment you introduce a meter, you change behaviour

Imagine two CTAs inside a product.

One says: Analyze this account.

The other says: Analyze this account. 47 credits.

The functionality is identical. The behaviour will not be. The moment users become conscious of consumption, they start making micro-economic decisions the product previously made irrelevant. Is this account important enough? Should I save the credits for later? Which teams get access? What happens if everyone starts using it?

Adoption now has an internal competitor: cost anxiety.

This is especially dangerous for AI, because customers need to experiment before they understand where the highest value lies. Early AI adoption is inherently exploratory. People try workflows, stumble into unexpected use cases, and gradually build conviction about where automation matters.

Consumption anxiety teaches the opposite behaviour: use AI only when you’re already sure the task is worth paying for.

Which creates a strange paradox. The pricing model designed to monetize AI suppresses the experimentation customers need to discover enough value to pay for it.

Credits price what the vendor understands, not what the customer values

There is a second structural problem with most credit systems. Customers do not wake up wanting tokens, inference seconds, or credits. Those are abstractions created by our architecture and our cost structure.

Customers want 5,000 calls reviewed. They want 300 accounts researched. They want the ten deals most at risk identified before Friday. They want three hours of repetitive work to become twenty minutes.

Now tell that customer: Your agent consumed 8,470 credits.

The obvious response is, “Was that good?”

This is where usage systems reveal their weakness. The product measures consumption with extraordinary precision while measuring value with almost none. The customer watches one meter fall and cannot see any meter rising.

When cost is precise but value is fuzzy, customers become conservative.

Never make customers understand your unit economics before they understand their own ROI.

AI pricing actually requires four kinds of trust

When companies talk about pricing transparency, they usually mean making the rate card visible. That matters. It is nowhere near enough. For consumption pricing to work, customers need four distinct forms of trust.

1. Unit trust: Do I understand what I’m buying?

A thousand credits means nothing to a business user. A thousand conversations analyzed means something. Five hundred accounts researched means something. The further the pricing unit sits from a job the customer understands, the more translation the customer has to perform. Every layer of translation creates friction.

The best pricing units speak the customer’s language, not the infrastructure’s.

2. Budget trust: Can I predict what this will cost?

Enterprise buyers dislike uncontrolled spend even when the ROI is attractive. Predictability itself has economic value. A CFO will approve a large predictable commitment more readily than a smaller one that could unexpectedly double.

AI makes this hard because consumption is not deterministic. A human opening a screen is easy to model; an agent deciding how much retrieval, reasoning, and tool use a task requires is not. Which means cost forecasting can no longer live inside Billing. If you introduce variable pricing, you inherit an obligation to help customers understand what drives the variable. Budgets, forecasts, and consumption simulations become part of the product experience.

3. Control trust: Can I stop the machine?

This matters far more as we move from copilots to agents. With traditional SaaS, a human had to be actively doing something to generate product activity. An agent can work continuously while nobody is sitting in front of the software. It can analyze data, invoke tools, and trigger actions in the background.

An administrator should never discover after the fact that an autonomous system has been autonomously spending their money.

If an AI agent can create economic activity, spend controls are product architecture, not a billing feature.

4. Value trust: Can I see what I received?

Imagine opening a dashboard and seeing only: 17,400 credits consumed.

Now imagine the same number alongside: 3,200 calls analyzed, 114 deal risks identified, 37 manager interventions triggered, 210 hours of manual work avoided.

The consumption hasn’t changed. The customer’s experience of it has. Serious AI dashboards eventually need two meters: one showing what I spent, one showing what I got.

The customer’s first surprise should be value, not a depleted meter.

So how should AI actually be priced?

There is no universal answer. “Everything will move to outcome-based pricing” sounds attractive, but different kinds of AI create value at very different distances from a measurable business outcome.

The better principle: price based on the nature of the work being performed.

Assistive AI should usually disappear into the product.

Summarize this conversation. Rewrite this email. Prepare me for this meeting. These are frequent, lightweight interactions whose strategic value comes less from the individual transaction and more from habit formation. Putting a visible meter on every interaction teaches users to avoid exactly the behaviour you’re trying to create.

Sometimes the smartest monetization decision is to make AI feel free, even when it is not free to you. The economic value shows up in platform price, retention, and differentiation rather than a line item.

Agentic work needs a unit of work customers understand.

When software performs substantial work autonomously, consumption pricing becomes easy to justify: the customer can see the software doing something that previously required real human effort. But even here there is a choice. Price the computation, or price the work completed.

Push toward the latter wherever possible.

Not 650 credits. Research 500 accounts.

Not 1.4 million tokens. Review 10,000 conversations.

Not 200 model invocations. Process 1,000 claims.

Priced this way, the vendor takes responsibility for efficiency. If better models, caching, or architecture let me complete the same task next year on half the compute, that becomes my margin improvement. I should never need to expose every evolution in my infrastructure economics to the customer.

Customers buy the job. We should own the machinery.

Move toward outcomes only when attribution supports it.

Outcome pricing appears to align the vendor completely with customer value, and in some categories it will be right. A support ticket resolved is measurable. An invoice recovered is measurable.

Charging for a closed enterprise deal is a very different proposition. Pricing, competition, salesperson quality, procurement, brand, and timing all determine whether that opportunity closes. This is why I resist the fashionable idea that every AI company should simply charge for outcomes.

Price as close to customer value as attribution allows, but no closer.

Good pricing requires economic courage. It also requires intellectual honesty about what your product actually caused.

Every AI pricing model is really allocating risk

This is where the conversation becomes bigger than pricing.

Flat subscriptions place usage risk on the vendor. Consumption pricing transfers some of it to the customer. Outcome pricing transfers performance risk back to the vendor. So choosing a pricing model is not choosing between seats, credits, or tasks.

It is deciding who carries the uncertainty.

And AI introduces a lot of it. How much will the customer use? How expensive will that usage be? How efficiently will models do this work two years from now? How attributable is the value? Every pricing architecture distributes those uncertainties differently between customer and vendor.

That is why pricing is ultimately a trust question. The more uncertainty we ask customers to carry, the more visibility, predictability, and control we owe them in return.

Pricing migrations are product migrations

One final implication for leaders.

When an existing AI capability moves from “included” to “metered,” this cannot be a Finance-and-Packaging launch supported by an FAQ and a new price sheet. It is a product migration. Customers need to see what they historically consumed, what that consumption looks like under the new model, and what value it created. They need tools to forecast usage, set budgets, and understand what happens at their limits. Pricing, Product, UX, Customer Success, and Finance all have to ship it together.

Because the customer is not merely experiencing a new way of paying.

They are experiencing a new relationship with the product.

The question isn’t “What should one AI credit cost?”

We are going to spend the next few years debating AI pricing: seats versus credits, tokens versus tasks, consumption versus outcomes. Those debates matter.

But there is a more fundamental question leaders should ask first:

What risk are we asking our customers to take, and have we earned enough trust for them to take it?

The winners in AI may not be the companies that invent the cleverest pricing metric. They may simply be the companies where customers continue to feel comfortable saying, “Use more.”

Because when your customer’s first reaction to successful AI adoption is to look for the off switch, you don’t have a monetization problem yet. You have a trust problem.

They call Accenture. Or Capgemini. Or both.

The pitch doesn’t lead with the outcome. It leads with the methodology. A framework. A named delivery approach. A team. The outcome is implied. The headcount is explicit.

The SOW gets negotiated over weeks. Phases - Discovery, Design, Build, Test, Deploy. Each phase has deliverables. Documents, mostly. A requirements specification. A technical design. A test plan. Nobody can yet define what “done” looks like for the actual business problem — but everyone can agree on what documents should exist at the end of each phase.

Behind the SOW sits the rate card. Every role priced. An onshore partner. A senior consultant. An offshore developer at a fraction of the rate. The engagement staffed to a peak team of 1200, split across London , Mumbai, Bangalore. The monthly invoice is a function of who showed up and for how long.

Change requests are where the real commercial action happens. Scope shifts, as it always does. Each change goes through a CR process — scoped, estimated in days, priced at the rate card, approved by a governance board. Ambiguity in the original SOW, which was somewhat inevitable, becomes revenue.

Testing alone sustains a team of 100 for six months. Manual scripts. Defect cycles. Three or four rounds before UAT. Every round billable. Nobody is incentivized to write software with fewer defects. The testing phase is a revenue line.

When the engagement ends, a support team stays on. Because the system is now complex enough that the bank can’t run it without the people who built it. The knowledge lives in the humans, not the system. By design or by accident , it doesn’t matter. The result is the same.

This is not a criticism of how those firms operated. It was a rational commercial model, executing perfectly with a logic.

The logic was simple: the more human effort a problem requires, the more valuable the engagement.

And a $300 billion industry is built on top of it.

Software services has moved through three distinct eras. Most incumbents are living in the second and mostly ignoring the third.

Era 1: Sell the people.

Infosys, TCS, Wipro built something genuinely remarkable. They took India’s engineering density, priced it at a fraction of Western labor costs, and created a delivery model so reliable that the Fortune 500 outsourced entire technology functions to it. The offshore pyramid was the innovation. Not the software. Not the methodology. The pyramid.

TCS doubled its revenue from $15B to $30B over the last decade. Its headcount nearly doubled too - from 320,000 to 608,000. Revenue per employee stayed flat at roughly $49,000.

That is not a technology business. That is a staffing business with excellent branding.

It worked perfectly - until the thing that made it work, human hours, stopped being the scarce input.

Era 2: Sell the solution

The smarter services firms saw what was coming and made a different bet. Stop selling people. Start selling solutions - proprietary frameworks, managed services, platforms wrapped around delivery. Charge for the capability, not the headcount behind it.

It worked. Infosys built Finacle. SAP and Oracle created ecosystems that entire industries ran on. Accenture moved up the value chain into managed services and industry-specific platforms. Margins improved. The pitch changed from “here are our people” to “here is our method.”

And it was genuinely prescient. Partially.

Because underneath every platform was still a pyramid. Implementation, configuration, change management - the translation layer between what the software could do and what the enterprise actually needed it to do. The last mile was always human. The billing just got more creative about hiding it.

The managed services contract replaced the SOW. But inside the managed services contract, people still showed up, did the work, and got counted. The commercial logic of Era 1 survived inside Era 2 - just wearing different branding.

Profitable. Still running. And not yet the full answer.

Because the last mile remained human. And human last miles have a cost structure that a different kind of company is about to make look very expensive.

Era 3: Sell the outcome.

The Era 3 company looks like a services firm from the outside. There’s a contract. There’s a scope. Someone is accountable when things go wrong.

But you never negotiate a rate card. You never approve a staffing plan. You never sign off on a phase document. You describe the problem, agree on what success looks like, and pay when it happens.

What runs underneath - agents, humans, some combination - is the vendor’s problem.

It is services, rebuilt from first principles. Priced on results. And structurally indifferent to how many humans it takes to get there - because the answer is increasingly: very few.

You send them the problem. They return the outcome. You pay for the outcome.

Anterior doesn’t sell hospitals software to manage prior authorization. It does the prior authorization for them — with agents, at software margins, against what used to be a billing line for a team of humans making phone calls. Graph AI didn’t pitch a pharmacovigilance platform to a frustrated enterprise client. It had a demo running in four weeks and a production rollout in eight — against an incumbent that had been working the same problem for eighteen months.

That’s Era 3. Not a tool that makes your team faster. An agent that makes your team optional.

Indian IT: the opportunity is larger than the threat. But only for some.

The Nifty IT index dropped 19% in eight trading sessions in early 2026. $50 billion in market value, gone. The fear: AI could automate core services revenue faster than the firms could pivot.

Vinod Khosla said India’s IT sector could almost completely disappear by 2030. The optimists inside TCS and Infosys said demand will expand, AI will create new categories, the industry has navigated every technology transition before.

Both are partially right. Both are missing the point.

The model is ending. The opportunity is enormous. These are not contradictory statements , they just apply to different companies.

The wave of Era 3 companies that disrupts Indian IT from below will not come from Accenture or from Silicon Valley. It will come from the people who spent five years inside TCS learning exactly which workflows are hollow and who now know how to fill them with agents instead of analysts.

India isn’t at risk of missing this wave. It is positioned to build it. The question is whether enough people inside the incumbents realize it before the incumbents find a way to make staying comfortable.

The pyramid doesn’t disappear. It inverts.

The AI-native services company doesn’t need to be small. It needs to be differently shaped.

The Era 1 pyramid had one senior person in the steering committee, a layer of architects beneath them, and a wide base of junior resources — developers, testers, BAs — doing the execution work offshore, billed at a fraction of the onshore rate but generating margin through volume. The pyramid worked because execution was expensive. Writing code, running test cycles, producing documentation — these required human hours, and human hours had to be staffed, managed, and billed.

That base is collapsing. Not disappearing , collapsing.

Agents now handle the execution layer faster, cheaper, and without a change request every time scope shifts. What used to require 200 people to deliver now requires 200 agents to run and 20 people to direct, evaluate, and own.

That is the ratio that’s changing. Not the output. Not the outcome. The humans required per unit of delivered work.

And what rises as the base compresses is not more of the same. It is a fundamentally different kind of human. The domain expert who knows where the agent breaks — specifically, not theoretically. The clinical nuance the prior auth agent misses in 2% of cases. The jurisdictional edge case in contract law the model consistently gets wrong. The AI engineer who can instrument whether it succeeded. The person who can price the outcome and close the contract on those terms.

And the person who picks up the phone when something goes wrong at a regulated institution and a human needs to be accountable.

A 200-person engagement becomes a 20-person engagement. A 20-person support team becomes a 3-person oversight function. The numbers compress at the base. The value concentrates at the top.

That is not a small change. That is the pyramid, turned upside down.

Why this is the most interesting space to build in right now.

The $300 billion IT services industry is mid-transition. Era 1 companies are training their workforces on AI tools and hoping the margin holds long enough for demand to recover. Era 2 companies have built real platforms and real recurring revenue — but still carry the human last-mile cost that Era 3 is about to eliminate. Era 3 barely exists yet — a handful of vertical-specific, agent-powered services companies, most under 3 years old, most operating in regulated industries where workflow complexity and accountability requirements are high enough to slow commoditization.

That gap will not stay open long.

The firms that enter it in the next two to three years have one structural advantage that cannot be replicated by incumbents: they will never have priced by the hour. They will instrument from day one. They will hire for domain depth, not delivery scale. They will build accountability into the product because they have no pyramid to hide behind.

The SOW was a document designed to make ambiguity safe for the services firm. Every change request, every phase extension, every support contract was ambiguity monetized.AI-native services companies will make ambiguity the client’s problem to define and the agent’s problem to resolve.

That is a fundamentally different commercial relationship. And it will produce a fundamentally different industry.

The UK bank engagement that took 1200 people and eighteen months in 2005 will be quoted differently soon.

Smaller team. Shorter timeline. No rate card. No change request process.

Just: here is the outcome. Here is what it costs. Here is who is accountable if it fails.

Are you seeing Era 3 companies emerge in your space? And are they coming from inside the incumbents, or from somewhere no one was watching?

A large enterprise customer - the kind whose logo you put on your website - had a straightforward ask. Their AI assistant was auto-generating weekly status updates for project teams. They wanted a setting to control the tone. Formal for executive stakeholders. Conversational for internal teams. Two options. A radio button. A one-day build.

The PM said yes. Of course she did. It was a reasonable ask from an important customer.

Then the next customer wanted a setting for length. Short for busy leaders. Detailed for working teams. Also reasonable. Also a one-day build. Then someone wanted control over which projects got surfaced automatically. Then another customer wanted the ability to suppress certain data points from the output depending on audience type.

Each request arrived wearing the same costume: small, reasonable, a few days of work.

Six months later, there were nineteen settings. The assistant was navigating hundreds of possible configuration states. And it had started behaving in ways nobody could fully explain. An update would be formal in one context, pull in the wrong projects in another, suppress a metric that a stakeholder had specifically asked to see. Every output was technically correct given the configuration. But to the end user, the product felt broken.

Support tickets increased. Trust eroded. The thing that had felt like the future now felt unreliable.

The PM hadn’t made a series of bad decisions. They had made very good B2B SaaS decisions. That was exactly the problem.

The instinct that built great SaaS products over the last decade - stay flexible, configure for every customer, never say no to a reasonable ask - is precisely the instinct that will break AI-native products.

Here is what actually happens when you add a setting: you are not adding flexibility. You are forking the product. Every new configuration state is a version of the product that must be tested, maintained, and proven to behave correctly. In traditional SaaS, a poorly considered setting created a confusing UI. In an AI-native product, it creates an unpredictable agent. And an unpredictable agent is a product killer in a way that a cluttered UI never was. Users will tolerate a complicated interface. They will not tolerate an AI they cannot trust.

But the underlying customer need is real. Enterprise teams do have different workflows. Different audiences require different outputs. The variation is legitimate.

What changes in the AI era is not whether variation exists. It is where the variation lives.

In traditional SaaS, variation lived in the product - in settings, configuration panels, and toggle states that PMs built and maintained. In AI-native products, variation should live in the intelligence layer. The agent adapts through context, memory, and natural language instructions , not through a UI that fragments into a thousand configurations the PM now owns forever.

Every setting you add is complexity. And complexity in an AI-native product doesn’t just slow things down — it compounds unpredictably.

That customer who wanted tone control? In an AI-native product, they should be able to instruct the agent directly: “For updates going to the executive team, always keep the tone formal and limit to three bullet points.” The agent understands. The agent remembers. The product stays simple for everyone else. The flexibility moves from the interface to the intelligence. The PM doesn’t build a setting. The surface stays clean.

This is the real design shift. Not less power. Not less flexibility. Just a fundamentally different place for it to live.

This plays out differently depending on where you are.

If you’re building AI-native from scratch, the pressure will come early. It will arrive in the form of a great customer relationship, a compelling renewal conversation, or a product review where someone says the words “it’s just one setting.” Resist it. Every time you feel the pull to configure something in the UI, ask first: can the agent handle this through instruction? Most of the time, the answer is yes. And when the answer is no, that is a signal worth paying attention to - it usually means there is a capability gap in the intelligence layer, not a missing setting in the interface.

If you’re replatforming an existing product onto AI, the pressure is worse. You will inherit a backlog of existing configurations and someone will ask you to port them across. Don’t treat replatforming as a migration. Treat it as a redesign. The settings your old product accumulated were workarounds for problems your old architecture couldn’t solve elegantly. Your new architecture can. Start from what the user actually needs, not from what the previous product had built up over years of saying yes.

The best AI-native products will feel almost unreasonably simple. Not because they do less. Because the complexity lives where users never have to touch it.

Simplicity isn’t the absence of power. It’s knowing where the power belongs.

And it doesn’t belong in a setting.

It is not a new idea. But it feels more urgent now , in a moment when we can build almost anything, when AI gives us velocity we did not have before, when the default answer to “should we add this?” has become a reflexive “yes.” Taste, in that context, becomes radically countercultural.

Here’s what I think taste actually is:

Sharp Choices

Taste is the ability to strip away all the noise and establish - first - what you are actually looking at.

What teams with taste do differently is they remove first, then build.

They ask: What is the bare bones version of this? But what is the essential thing stripped of every assumption, every “nice to have,” every thing that seemed obvious because competitors had it?

This is brutal work. It requires saying no to things that are technically sound, commercially viable, and probably requested by customers. But if they are not part of the bare bones, they are noise.

The distinction matters because most decisions that water down a product are not wrong decisions individually. They are just not essential decisions. And essential is the bar that matters.

Teams with taste know their core. They have stripped it down so completely that they can articulate it in one sentence. Then every feature that ships either serves that core or gets rejected, not because it is bad, but because it is not that thing.

Simplicity as Strategy

Simplicity is not about minimalism for its own sake. It is about removing everything that does not serve the core job.

The distinction matters because true simplicity is hard. It is not just “fewer buttons.” It is about understanding your users so deeply that you know exactly what they need to see, in what order, to accomplish their goal. It is about coherence, every element reinforcing the same thesis.

A simple product feels intentional. Users notice. They feel respected.

One Problem, Solved With Focus

This is perhaps the most important principle. Taste means picking a problem worth solving and solving it better than anyone else.

The temptation to expand is constant. Adjacent features call out. New markets open up. But products with taste stay disciplined. They get deeper, not wider. They refine, not sprawl.

The Layers Beneath the Surface

But there’s more to product taste than these three pillars. The best products reflect other choices too:

Clarity over cleverness. A product with taste doesn’t require users to decode it. The why is obvious. The mental model is intuitive. There is no mystery or assumed context. This matters everywhere, but especially in B2B where users are time-constrained and have seen a thousand products promise more than they deliver.

Opinionated defaults. Rather than maximizing optionality, taste means shipping with strong defaults that reflect your point of view about the right way to solve this problem. The user can deviate if they need to, but the baseline says: “This is what we believe. This is how it should work.” It’s confidence expressed through design.

Speed to value. There is an underrated dimension here: products with taste deliver ROI fast. They respect the user’s time by getting them to value in days, not months. They don’t require extensive configuration or tribal knowledge. This is especially true for AI-first products - the products people will actually adopt are the ones that prove their worth quickly, not the ones with the most comprehensive feature set.

Coherence across the entire experience. Taste isn’t just about one layer, it’s woven through the entire product. The pricing reflects the positioning. The onboarding teaches the right mental model. The feature set doesn’t argue with itself. When you ship a product with taste, you notice the absence of contradictions.

The Real Test

The hardest part about shipping products with taste isn’t the design thinking. It’s not the strategy conversations. It’s the discipline required even when you have the engineering velocity to say yes.

Let’s say hypothetically, every week, your team can probably build the feature request that just came in. Assume you have the capacity. Assume you have the technical chops. Assume that the answer is - “always possible”.

Taste means saying no anyway. Not “no, not yet.” Just no.

That is what separates products that matter from products that don’t.

Because velocity without taste is just noise at higher speed. It’s more features, faster. More options, with less clarity about what you’re actually building.

The products that will define the next era won’t be the ones with the most capabilities. They’ll be the ones with the clearest sense of purpose. The ones that solved one problem so well, so cleanly, that users couldn’t imagine doing it any other way. That is Taste. And it has never been more valuable.

What does product taste mean to you? I’d be curious to hear where your definition differs or where you’ve seen it matter most.

And in the middle of all this change, one thing hasn’t moved: Strong product fundamentals still matter. They always will. They’re the foundation of every competency model I’ve built and every team I’ve led.

But in the AI era, fundamentals aren’t the only differentiator anymore.
Spikes are. Team composition is.

When I say spikes, I mean the places where someone is undeniably world-class.
The strengths that are so sharp, so natural, so instinctive that they change the velocity of everyone around them.

A spike is:

• the person who senses a platform shift before anyone else

• the person who can align 40 people with one narrative

• the person who pulls truth out of noisy data

• the person who ships through ambiguity without losing momentum

You don’t hire spikes for balance. You hire spikes because their strength is so strong it creates new possibilities.

That’s why team composition matters more than individual completeness.
One big spike paired with another big spike is how modern product orgs unlock disproportionate impact.

And different people instinctively gravitate toward different parts of this new world.

Some light up when new agent patterns drop.
Some thrive when building teams and setting long-term direction.
Some find truth in data chaos.
Some are relentless finishers who ship through ambiguity.

And the magic happens when these strengths combine. A portfolio of differentiated strengths that slot together like a great football team: one striker, one playmaker, one controller, one defender.

Here are the six archetypes I now believe every modern SaaS product org needs.

Disclaimer - I’m not calling out “customer value” as a separate archetype. That’s intentional. Customer understanding, empathy, and insight are the baseline expectation of every product role. It’s the reason the function exists. Everything else sits on top of that foundation.

Also people may overlap across 1 or more spikes, but will always have a strong core strength

1. The Tech Visionary

The person who naturally lives 18 months in the future.

They are the kind who will tear apart OpenAI’s new release the same night it drops.
They are reading pricing appendix notes.
They are reverse engineering agent frameworks.
They are building side prototypes at 2 AM because a new API landed.

In the AI world, this archetype is no longer “nice to have.” They are your early-warning radar system.

What they bring:

• Anticipate platform shifts before the org feels them

• Spot real inflection points vs hype cycles

• Understand how model capabilities unlock new workflows

• Build zero-to-one proofs of possibility

• Challenge “this is how we do it” with “this is how it will be done”

Where they struggle:

• Quarterly planning

• Anything that involves slow consensus

• Anything that smells rigid process

Use them like an internal R&D lab, plugged directly into customer pain.
Give them air. Give them freedom.

Without this person, your roadmap slowly becomes a museum.

2. The People + Product Builder

Your Org’s Orbit Stabilizer

Every org has one gravitational force. The one who keeps everyone aligned, engaged, pointed in the same direction.

This person is half coach, half storyteller, half organizational architect.
Yes, that’s three halves but you know the type.

What they bring:

• They design teams

• They shape culture

• They build product narratives that unlock belief

• They energize people

• They connect the dots across strategy, craft, and execution

• They create rituals that turn chaos into cadence

AI makes this archetype even more necessary. Why?

Because when tools move faster than people, people need anchors.
They need clarity, direction, and a sense of meaning.
They need someone who says, “Here is the world we are building… and here’s your role in shaping it.”

Where they struggle:

• Deep technical immersion

• Fast prototyping

• Getting hands dirty every day

Use them to run product areas, cross-functional pods, org design, and narratives.
They are the stabilizer.They convert disruption into direction.

3. The Data Decoder

The AI world produces more data than teams know what to do with. Usage analytics, embeddings, signals, drift metrics, activation loops, cross-system traces…
It’s noise unless someone can translate it.

This archetype is your truth-seeker.

What they bring:

• Causal models, not dashboards

• Pricing and packaging sensitivity analysis

• Activation and retention diagnostics

• Early signal detection (what’s trending up or collapsing)

• Ability to debunk magical thinking

In an AI-first product, this PM becomes even more valuable:

• They evaluate whether agents are actually improving outcomes

• They validate real vs perceived value

• They pressure-test hallucinated narratives

• They build the evidence for strategy, not the spreadsheet for a deck

Where they struggle:

• Emotional storytelling

• Messy frontline conversations

Deploy them in growth loops, monetization, enablement, and workflows with scale.
They are the BS filter of the organization.

4. The Execution Engine

The person who quietly ships more work in a quarter than most teams ship in a year.

Every org has one.The person who moves through ambiguity like a heat-seeking missile. No ego, no drama, no paralysis. Just relentless momentum.

What they bring:

• Roadmap throughput

• Coordination superpowers

• Decision velocity

• Ability to lead engineers, design, and CS in one thread

• Crystal-clear follow-through

In AI products, where half the work involves ambiguous APIs, shifting infra, and cross-functional unknowns, this person is gold.

Where they struggle:

• Long-term strategy

• Models and frameworks

• Category creation

They are the engine that converts strategy into impact.

5. The Product Architect

The Systems + COGS Thinker

And in AI-heavy systems, there’s a fifth spike that becomes crucial.This is the product person who thinks in building blocks, not features. They understand how the entire system fits together — the data layer, the workflow layer, the AI layer, the cost layer. They have a mental model of the product that sits somewhere between a strategist and a platform engineer.

If the Tech Visionary sees what could be built, the Architect PM sees how it should be built without breaking cost, coherence, or the future.

What they bring:

• Systems thinking across workflows, APIs, and AI agents

• Ability to break problems down into durable, reusable building blocks

• Deep awareness of long-term COGS: inference cost, storage, retrieval, context windows, latency

• Judgment on sequencing: what needs to exist first for future speed

• Intuition for scale, interoperability, and architectural consequences

In the AI era, this spike becomes critical. Agents, embeddings, orchestration, caching, and retrieval all create hidden complexity.
Someone needs to say, “If we do this, here’s what it does to cost, latency, and maintainability.”

Where they struggle:

• Moving fast when trade-offs aren’t resolved

• Simplifying for stakeholders who want quick narratives

• Anything that feels like “just ship it”

Where to deploy them:

• AI architecture and COGS strategy

• Reusable frameworks and building blocks

• New product foundations

They are the safeguard against spaghetti systems.They protect the future of the product.

6. The Experience Architect

The keeper of invisible UX.

AI introduces power but also chaos. Left unchecked, you end up with prompts everywhere, random flows, inconsistent agent behavior, cognitive overload.

This person keeps the product calm, clean, and inevitable.

What they bring:

• Obsession with simplicity

• Invisible guidance

• Predictable agent interactions

• Reduction of cognitive load

• Coherent system-wide UX patterns

• The instinct to hide complexity, not expose it

They ask the questions that save users:
“Why does the user need to see this?”
“Can we collapse these steps?”
“Does this feel like work?”

This isn’t a PM role. Not just design either.
It’s a craft spike - the experience conscience of your product org.

The Real Unlock:

You don’t build an AI-era product org by hiring similar people.
You build it by composing complementary spikes.

The right combinations change everything:

0→1 AI Bet

Tech Visionary + Data Decoder + Experience Architect

Scale Phase

Org Builder + Execution Engine + Experience Architect

Replatforming for AI

Tech Visionary + Org Builder + Execution Engine

Pricing, Growth, and Expansion

Product Architect + Data Decoder + Product Builder

Teams fail not because the people are weak. They fail because the portfolio assignment is wrong. The future of product leadership is about assembling spike-stacks. It’s about pairing the dreamer with the builder, the doer with the analyst, the visionary with the stabilizer.

In an AI-first world where velocity is absurd, uncertainty is constant, and the ground shifts every few quarter, the composition of your team matters more than the capabilities of any one person. You win by composing a team where every person bends the curve in their own way. Because in a world where the ground shifts every few quarters, the real competitive advantage isn’t individual brilliance.
It’s the architecture of the team behind it!

Every day, I see how technology can personalize learning for adults - understanding how they think, where they struggle, what keeps them engaged. As a parent of a 14-year-old, I see the opposite at school.A one-size-fits-all system that measures everything except what truly matters.

I see it every week - the small, quiet absurdities of our school system.

• My son spends hours memorizing definitions he will forget by next month.

• In math, he’s told to copy steps exactly as taught - even when he finds a quicker way to solve it.

• In science, experiments are pre-written, results pre-decided, and curiosity optional.

• History rewards rote recall of dates and dynasties, not an understanding of why civilizations rose or fell.

He’s not graded on originality of thought. The report card that lands in my inbox every term tells me almost nothing about who he’s becoming , only how well he can reproduce what’s been taught. Everything is measured, but almost nothing is discovered.

The majority of ICSE and CBSE schools still revolve around the same formula:textbooks, tuitions, tests, and toppers. Marks are the only language that matters.We do have a few progressive schools - ones that blend art with science, teach through projects, and let curiosity lead. But they remain rare and mostly unaffordable for a vast majority.

As a parent, it’s frustrating. As a product leader who builds AI systems for complex human problems, it’s alarming. This made sense when the goal of education was to produce efficient workers - people who could memorize, follow, and reproduce.That system worked for the industrial age.It even worked for the service economy. But it will not work for the world that’s coming next.

My son’s generation will grow up in a world that looks nothing like the one we prepared them for. Because the world is changing faster than schools are evolving. Today’s world rewards synthesis, not memorization. Judgment, not compliance. Curiosity, not quiet obedience.

And here is the unsettling truth - we have no idea what most careers will look like 20 years from now.Every major industry is being rewritten. The safest skill might just be learning how to learn.

Our classrooms were designed for certainty - same lessons, same tests, same definitions of success.But the future belongs to those who can connect ideas, think critically, and keep learning as the world changes. Can AI help schools move from repetition to creation and finally teach children how to learn?

So What Needs to Change

Three things.

Curriculum.
Teaching.
Assessment.

Our curriculum is still built like an assembly line. Math here, history there, biology somewhere else. But real life does not work in silos.

When a pandemic hits, you need biology to understand the virus, math to model its spread, economics to manage the fallout, and ethics to make the hard calls.
But you also need behavioral science to understand why people resist masks, data visualization to track trends, and history to remember how societies handled plagues before. When we teach these things separately, students learn facts, not frameworks.

It is the same with climate change that is taught as geography, not as physics, politics, or human behavior.

Or with AI reduced to coding, when it is equally about philosophy, law, and ethics. Even something as simple as urban traffic is a systems problem: math for flow, psychology for decision-making, policy for regulation, and design for safety.

Real life sits at the intersection. Our classrooms still sit in silos.

A modern curriculum should show how physics meets philosophy, how biology and data science explain health systems, how history and politics predict the next wave of change. It should make kids connect ideas - not memorize them.

Teaching, too, needs a complete rethink. The best teachers are not information givers; they are curiosity engineers.

Think of how a Sandeman’s walking tour guide walks you through Barcelona - not by listing dates, but by weaving stories that make you feel the city.Or how a YouTube creator like Veritasium starts every lesson with a mystery, not a formula. That’s what real teaching feels like - pulling thinking out, not pouring facts in.Bloom’s old pyramid put “remembering” at the bottom and “creating” at the top.Most of our classrooms still live at the bottom.

And then comes assessment - the most outdated part of all. We still test who can remember best under stress.But the world doesn’t reward memory anymore; it rewards reasoning, creativity, and the ability to adapt.

Imagine if students were asked not to write about different climate zones on earth but to design a sustainability plan for their own city. That’s how you test understanding - by asking them to use what they know.

We still ask every child to climb the same tree. But an elephant will fail. So will a fish.

Each child has a different kind of intelligence - spatial, analytical, emotional, artistic.
Our job isn’t to standardize them. It’s to surface what they’re wired for.

For the first time in human history, AI can make that possible. Not as a shortcut, but as a mirror - reflecting how each child learns, thinks, and struggles.

AI can observe subtle patterns that humans miss: Which students learn best through visuals? Who thrives in peer debate? Who shows persistence in projects but freezes in tests?

It can personalize learning paths at a level no traditional system could. It can help teachers focus on mentorship instead of monitoring. And it can give schools the tools to adapt - fast, fairly, and at scale.

This is the real opportunity: to use AI not to replace teachers, but to amplify them.

As a product leader working at the intersection of learning and AI, I’ve seen this firsthand. AI doesn’t create transformation on its own. It creates leverage - the multiplier that lets humans do what they already know is right, but at a scale and precision never possible before.

The School System Opportunity Landscape

Most edtech innovation so far has orbited around schools - not inside them.
Apps for parents, after-school tutoring, adaptive tests. But the real leverage lies within the school walls. Because that’s where a child spends most of their waking hours.
And that’s where transformation has to take root.

Here’s where founders, educators, and policymakers can focus - to make schools more human, more adaptive, and more future-ready.

1. The AI Teaching Assistant — every teacher’s invisible ally

Every teacher faces the impossible: thirty different learners, one lesson plan.
An AI co-teacher could change that. It listens, analyzes engagement, and suggests adaptive prompts or analogies in real time. It tracks who’s drifting off, who’s curious, who’s struggling - and quietly supports the teacher with context.

It does not judge. It amplifies awareness.

Why it matters: Teachers are the bottleneck and the multiplier. AI gives them leverage, not replacement.
Force multiplier: turns one teacher into a mentor for thirty unique learning journeys.

2. The Dynamic Curriculum Engine

A modern curriculum shouldn’t be a list of subjects. It should be a network of problems worth solving. Each problem - whether it’s climate, misinformation, or AI ethics - draws from multiple disciplines, forcing students to connect dots instead of memorizing them.

Here’s what that could look like inside a school redesigned for the future

The Math of Misinformation

Subjects: Mathematics · Psychology · Media Studies · Computer Science

Project:
Students trace how a rumor or viral post spreads through social network, model its reach using probability, and design interventions - fact-checking tools, content labels, or awareness campaigns - to slow its spread.

What it teaches:

• How statistics and data visualization can expose bias.

• How psychology explains why people believe what confirms their worldview.

• How algorithms shape collective behavior.

Real takeaway: Truth has math behind it — and so does deception.

The Economics of Attention

Subjects: Economics · Psychology · Design · Computer Science

Project:
Students analyze their own screen-time data, map the triggers that keep them scrolling, and prototype a “humane” app that rewards focused engagement instead of distraction.

What it teaches:

• Economics of attention as a currency.

• Psychology of habit formation and dopamine loops.

• Design and computer science for ethical interfaces and algorithms.

Real takeaway: The product is free because the user’s attention is not.

The Physics of Imagination

Subjects: Physics · Art · Philosophy · History of Science

Project:
Students turn invisible scientific concepts into sensory experiences — painting what time dilation feels like, or animating the birth of a star. The goal: explain one physics concept through art and storytelling.

What it teaches:

• Physics of motion, energy, and time.

• Art and philosophy to visualize the abstract.

• History of science showing how imagination precedes discovery.

Real takeaway: Creativity is not the opposite of science - it is the engine of it.

Why it matters: Teachers want flexibility but lack time. AI gives them creative superpowers.
Force multiplier: compresses curriculum design from months to hours while keeping teacher control.

3. The Continuous Assessment Platform

Replace one-shot exams with ongoing, evidence-based evaluation. Students upload projects, prototypes, and reflections. AI helps teachers tag, comment, and visualize growth over time. The data shifts from “who topped” to “who improved,” from marks to mastery.

Why it matters: What we measure is what we value.
Force multiplier: makes fair, qualitative assessment scalable across hundreds of classrooms.

4. The Teacher Growth Studio

If teaching must change, teacher learning must too. This platform models the pedagogy it preaches: collaborative, reflective, interdisciplinary. Teachers learn through micro-courses, share classroom experiments, and co-create new learning modules. AI curates mentors, highlights best practices, and builds a dynamic learning community.

Why it matters: Every education reform depends on teachers who can grow with it.
Force multiplier: democratizes professional development and builds peer-driven culture.

5. The Learning Record Infrastructure

The invisible backbone of a reimagined school system. A secure, longitudinal record of every student’s growth - projects, feedback, and reflections - across years.
AI analyzes this data ethically to identify strengths, learning gaps, and potential pathways.

This becomes a lifelong Learning Passport - portable across schools and eventually across careers.

Why it matters: The world won’t be credentialed by marksheets, but by evidence of capability.
Force multiplier: makes authentic learning visible, measurable, and mobile.

AI is not the goal. It is the accelerator - the multiplier that makes scale, personalization, and equity possible at once. This is not about replacing schools with apps. It’s about retooling schools themselves - curriculum, teaching, and assessment with AI as the quiet infrastructure underneath.

As a product person building AI-first systems, I know this isn’t just a technology problem , it’s a design problem. And as a parent, I know it’s not optional.
The next generation deserves schools that evolve as fast as they do.

The last century proved that real breakthroughs happen not in comfort, but in crisis.
World War II forced humanity to learn differently - and that changed everything: radar, rockets, computing, penicillin.We are at a similar inflection point , not of war, but of possibility. We can keep producing people for jobs that no longer exist, or we can raise a generation of creators, problem-solvers, and thinkers who will build what comes next.

I don’t think this can be fixed with workshops or pilots. It’s not a curriculum update we need - it’s a complete reset.

Because what we call “school” was built for predictability and deterministic outcomes. And predictability is gone.

We cannot prepare children for the future using the logic of the past. We need a system that’s not reformed - but rebuilt. From first principles.

It will take something more powerful than policy - it will take a movement.
Led by educators who refuse to teach inside boxes.By parents who stop mistaking marks for mastery.By founders who see schools not as a market, but as humanity’s most important infrastructure.And by technologists who design AI not to instruct, but to illuminate - to make learning visible, personal, alive.

This won’t happen gradually.It will start where pressure meets possibility - the schools that are bold enough to unlearn before they teach.

And when it comes, it won’t feel like change. It will feel like the world finally catching up to what every child already knows - that curiosity, not compliance, is the real measure of intelligence.

Moab was not on my calendar when I landed in US. A few meetings in Salt Lake City took us out of Chicago and landed us into a beautiful city close to some spectacular national parks . In a last minute decision - we got ourselves a rented car, and that impulse led us into a stretch of desert that kept unfolding surprises at every bend.

We drove into Arches first - vast red walls curved against the sky, each formation a patient reminder that endurance can be beautiful.

Then Canyonlands, where the horizon bent into layers of red and ochre, each line marking time I couldn’t fathom.

Two hundred million years ago, this land sat under an inland sea. Layers of sandstone, silt, and salt built upon one another, then buckled, cracked, and weathered away. Wind and water did their patient work - carving what remains into bridges of stone and canyons so vast they seem to hold time itself.

I had no expectations - no checklist of viewpoints to capture. I thought I’d find beauty, but I wasn’t prepared for awe. Moab didn’t just impress me; it transported me. Standing there felt like being pulled back in time — not decades, but millions of years.

And then we went up Hell’s Revenge , a trail that lives up to its name. The 4x4 tilted at impossible angles, the tires gripping slickrock that looked like it belonged on another planet. Heart pounding, sand in our teeth, the wind loud in our ears. It wasn’t danger that thrilled me, it was the rawness , and how alive the landscape made me feel, how it demanded both trust and surrender.

I went in with no plan, and that turned out to be the point.

Standing in Moab, you can’t escape the scale of time. Millions of years compressed into a cliff face, an arch that looks effortless but is anything but. Erosion did what ambition never could: shape permanence out of patience.

It made me slow down. Not in the tourist sense of “switching off,” but in the sense of realizing that much of what I run after operates on timelines so short, they shrink to irrelevance when placed against a desert wall. The desert keeps time differently. And for a few days, so did I.

I’ve been thinking a lot about growth , about what changed in the last half a decade: the beliefs I held, the ones I let go of, the relationships that strengthened and frayed. Somehow in Moab, those thoughts felt louder. Not because the desert spoke, but because it didn’t.

Also , there is a parallel I can’t ignore.At work, we prize speed and planning, though much of it unfolds in unpredictable chaos.But the things that endure - trust, clarity, culture, even real innovation - they don’t sprint into existence. They are eroded into being. Pressure, time, persistence.An arch is as much what’s missing as what remains.And maybe that’s true of us too. I didn’t find answers in Moab. No “aha moment,” no revelation that rewires the next decade of my life. What I left with were better questions.

Questions about growth - not the kind you can present on a slide, but the kind that shows up quietly. In how you breathe. In how well you sleep. In how you respond instead of react, or how at ease you feel in your own skin. In whether your body feels like an ally, your mind feels uncluttered, and your choices feel deliberate. For years, I treated health as something to manage; lately, it feels more like the foundation I had forgotten I was standing on.

And the truest measure of growth might simply be how grounded I feel within myself. The steadiness in discomfort. The grace in letting go. The calm in not having to prove anything.

Questions about relationships - the ones that stretch beyond years or roles you play. I’ve started noticing how different conversations feel when I’m not half elsewhere, when I’m not preloading my next sentence. Presence has its own gravity. It deepens what’s already there. It turns time with people into something that leaves a trace.

I’ve found closeness with people I once thought I’d only know professionally, or in passing. Depth that arrived quietly, through small, unplanned moments - shared humor, vulnerability, listening without agenda. It reminded me that connection doesn’t always grow where you invest most effort; sometimes it finds you when you finally slow down enough to let it.

Questions about achievement - I’ve been fortunate to experience so much of what I once defined as success. Each milestone felt meaningful in its time, yet I now find myself drawn to a different question: what does achievement look like when it’s no longer about reaching, but about resonance? About creating things that move people, not just metrics.

Achievement, I realized, isn’t something to chase - it’s something to grow into. It’s not a summit but a state. Less about speed, more about substance. Less about what gets built, more about what endures.

Happiness is a target; contentment is a state.
Happiness says, “I’ll feel fulfilled when…”
Contentment says, “I’m at peace with what is, even as I grow.” Contentment is deeper.

It comes from learning to simply sit. To notice. To be aware without always becoming something. When you are present , truly present , time doesn’t slip through your fingers. Each moment becomes distinct, felt, lived.

Contentment comes from:

• Being present without proving anything

• Building because it’s interesting, not impressive

• Conversations without agendas

• Listening to my son explain a new idea, argue a point, or lose himself in a book — and realizing that these are the moments that define a life

• Doing work that feels aligned with who I am, not just what I’m good at

• Finding steadiness in uncertainty, and gratitude without occasion

• Letting ambition coexist with ease

Maybe that’s what Moab gave me - not answers, but orientation.
A reminder that stillness isn’t the opposite of motion; it’s what makes motion meaningful. That even as I build and lead and move fast again, I can hold onto this slower rhythm beneath it all - the awareness that not everything valuable is measurable, and not every horizon needs to be chased.

I’ve spent much of my life in forward motion — solving, scaling, shaping things that grow fast. This experience reminded that I also need to tend to the things that grow slowly — my health, my relationships, parenting, my curiosity, my craft, and the kind of wisdom that comes only when you stop trying to hurry it.

Some things like clarity of strategy, a product philosophy, or a life that feels whole — are built the same way arches are: slowly, quietly, by what time takes away.

And back in Moab - The night sky never cleared. The stars stayed hidden.
But maybe that was the point.
Sometimes the sky doesn’t have to reveal itself - you have to see it for yourself.

But here’s the uncomfortable question no one wants to answer:

Are they actually doing any of this well? And more importantly — how would you know?

We have become brilliant at building with LLMs. We are still learning the ropes on judging them. What we lack is a trusted, repeatable, transparent system for measuring model performance where it matters most: in real-world use.

This gap is where LLM evals live. And it is why they matter more now than ever before.LLM evals — short for Large Language Model evaluations — are how we move from “it looks good” to “we know it works.”

At their core, LLM evals are tests. But not the kind we’re used to in software engineering. You’re not checking if the model compiles or throws errors. You’re asking deeper, fuzzier, high-stakes questions:

• Is the response factually correct?

• Is it helpful in context?

• Is it consistent, fair, safe?

• Does it align with your brand, your goals, your policies?

These aren’t easy things to measure. But they are essential.

Because here’s the thing about LLMs: they’re not like traditional software. They’re stochastic. Non-deterministic. Two identical prompts can yield subtly (or wildly) different results. And those results can change week to week — even day to day — as providers silently update models behind the scenes.

Which means that without evaluation, you’re not iterating. You’re just guessing.

The New LLM Deployment Stack: Evals at the Center

In the early days of AI, the excitement was about capability. Now, the excitement is turning into caution. Because in real-world deployment, a good demo isn’t enough.

You need systems to measure, compare, and improve model behavior across versions, vendors, and use cases.

That’s where LLM evals come in.

They act as:

• The QA layer in your LLM stack

• The trust engine for your AI product

• The flashlight when something breaks in production

• The scoreboard when you need to compare GPT-4 vs Claude vs Mistral

Use Case: A Sales Email Generator

You’ve built a GPT-powered tool that takes product details and customer pain points, and turns them into outreach emails for sales reps.Benchmarking with BLEU or ROUGE scores won’t help much. Those were designed for machine translation. They don’t capture:

• Is the tone persuasive?

• Is the content relevant to the persona?

• Is the CTA clear?

What to measure instead:

• Brand tone match (via classifiers or human grading)

• Persona-fit (was the email tailored to a VP of Finance vs a Sales Manager?)

• Call-to-action strength

• Repetition or hallucination of details

Use Case: Internal Knowledge Retrieval (RAG)

You’ve hooked your LLM up to internal documents via retrieval-augmented generation (RAG). When a customer success rep asks, “What’s the escalation policy for our Tier 1 accounts?”, the model fetches the relevant docs and drafts an answer.

Here, the eval needs to test both retrieval and reasoning:

Metrics to consider:

• Retrieval fidelity (did it pull the right doc?)

• Groundedness (is the answer based on retrieved facts?)

• Coverage (did it skip relevant documents?)

• Factual correctness, contradiction rate

From Static Benchmarks to Scenario-Based Evals

Open-source benchmarks like MMLU, GSM8K, and MT-Bench are useful sanity checks. They tell you if your model understands math, logic, or common sense.

But they’re not enough.

The future of evals is scenario-based:
Testing your model on tasks your users will actually do.

These include:

• Multi-turn conversations

• Real documents and messy edge cases

• Policy enforcement

• Domain-specific language

• Brand tone adherence

The Infrastructure Gap — and Who’s Filling It

As LLMs shift from prototypes to production, teams are realizing they need not just evaluation theory but real infrastructure — tooling to manage this end-to-end.

Here’s who’s leading the charge:

• Maxim: Scenario-based evals-as-code with regression testing, version control, diffing, and pass/fail analytics.

• Gaia: Multi-turn agentic evaluation using simulated user personas.

• LangSmith, PromptLayer: Observability tools with support for in-production evals and LLM output tracking.

We’ve been using Maxim, and it has mature approach to making LLMs predictable, testable, and safe.

Inside Maxim: Turning LLM Behavior into a Testable System

In most modern AI stacks, the interface layer is polished, the foundation models are powerful, and the orchestration logic is improving. But the middle layer — the layer responsible for evaluating and ensuring consistent behavior across prompts, updates, and use cases — is often missing or immature.

Maxim is building that missing layer.

It treats LLM behavior not as a mystery to observe post hoc, but as a system to actively test, monitor, and compare. Think of it as a behavioral test framework for AI — one that brings the discipline of software QA into the probabilistic world of language models.

What Maxim Solves

Most teams today rely on spot checks, static benchmarks, or subjective human reviews to determine whether an LLM “works.” This does not scale. Worse, it fails silently when models drift, APIs update, or prompt chains change.

Maxim provides a structured, repeatable way to evaluate model behavior across:

• Model upgrades (e.g., GPT-4 → GPT-4 Turbo)

• Prompt or system message changes

• Retrieval pipeline adjustments (RAG)

• Fine-tuning iterations or vendor shifts

Instead of relying on anecdotes or “vibe-based QA,” teams can codify expectations and run comprehensive evals every time something changes.

Scenario-Based Evals, Not Abstract Benchmarks

Maxim allows teams to create task-specific test suites tailored to actual use cases. These are not toy prompts or synthetic quizzes — they are real user flows, queries, and business-critical interactions.

• Define prompt + context (e.g., “A support rep asks about refund eligibility for a damaged product in EMEA”)

• Set reference outputs or evaluation criteria (e.g., must cite policy doc, avoid over-promising)

• Execute across one or more models and compare behavior

This makes it ideal for domain-specific, high-stakes applications where public benchmarks offer little signal.

Flexible Evaluation Modes

Maxim supports a variety of scoring mechanisms that can be layered or customized depending on task sensitivity:

• LLM-as-judge: Use GPT or Claude to compare outputs against references using semantic similarity, factual consistency, or rubric-based reasoning.

• Rule-based checks: Enforce hard constraints such as “must include CTA,” “no discount offered,” or “cite source doc.”

• Human-in-the-loop scoring: Integrate expert human reviewers for subjective or high-risk categories (e.g., legal summaries, healthcare advice).

This flexibility allows teams to optimize for both automation and accuracy, choosing the right evaluation fidelity for each use case.

Cross-Version Diffing and Regression Testing

The true power of Maxim lies in its ability to track changes across time — across model versions, prompt updates, or pipeline shifts.

With built-in support for:

• Version-controlled test suites

• Side-by-side output comparisons

• Regression summaries and dashboards

…teams can instantly see where performance improved, where it degraded, and whether any new failure modes were introduced. It replaces “model X seems better” with data-driven answers like “model X improved on factual accuracy by 18% but regressed on tone compliance in financial services use cases.”

This is essential for teams deploying LLMs in production, especially those governed by compliance, legal, or reputational risk.

Metric Aggregation and Behavioral Analytics

Each test suite run produces granular and aggregate metrics, which can be rolled up by use case, intent type, model version, or scoring dimension. For example:

• Task completion rate (per category or persona)

• Hallucination detection (based on groundedness to source)

• Pass/fail summary by model type

• Drift patterns over the past 30 days

This turns qualitative behavior into quantifiable signals that product, engineering, and risk teams can align on. And because these are embedded into the CI/CD process, evaluation becomes a continuous feedback loop, not a one-time QA checkpoint.

Maxim doesn’t replace your orchestration or prompting logic — it complements and hardens it. Think of it as the observability and quality assurance layer in a production-grade AI system.

It sits between:

• Your orchestration layer (e.g., LangChain, LlamaIndex, custom agents)

• Your model APIs or endpoints (e.g., OpenAI, Claude, Mistral, fine-tuned models)

• Your end-user apps (e.g., AI copilots, bots, assistants, or workflows)

Maxim becomes the single source of truth for evaluating model behavior over time, across multiple systems, use cases, and configurations.

Enterprise Use Case: AI Sales Copilot

Imagine a large B2B SaaS company deploying an AI assistant that helps sales reps draft outreach emails, handle objections, and respond to RFPs.

Without Maxim:

• A model change may cause tone to drift from formal to casual — unnoticed.

• A fine-tuning update may accidentally reintroduce hallucinated competitor comparisons.

• An update to the retrieval corpus may silently reduce accuracy for EMEA use cases.

With Maxim:

• A comprehensive test suite flags tone drift, hallucination, and coverage regressions.

• Sales leadership receives confidence scores by industry vertical.

• Product teams can safely upgrade the model without worrying about unanticipated fallout.

This isn’t just evaluation — it’s risk management and trust assurance, embedded directly into your development workflow.

Why tools like Maxim Represents a Strategic Shift

Maxim is not just a tooling innovation. It reflects a new mindset: that LLM behavior is not static, and trust is not a one-time decision.

In a world where models are probabilistic, APIs change without notice, and prompts evolve weekly, the only way to build resilient systems is to evaluate continuously, contextually, and systematically.

Tools like Maxim make that possible.

It moves LLMs from a “black box with occasional surprises” to a well-instrumented system that teams can trust, iterate on, and scale.For AI-native teams building real products — not just demos — this is no longer a nice-to-have.
It’s the foundation for everything else.

How to Build Your Own Eval System

You don’t need to be OpenAI to build a world-class eval layer. Here’s how to get started:

1. Define What “Good” Means for Your Use Case

Before you test anything, align on success criteria.
For your AI legal assistant, is “good” about completeness? Brevity? Risk minimization?

For your marketing copywriter, is it tone, clarity, or creativity?

Tip: Create a shared rubric with your stakeholders (product, legal, ops) before you write any code.

2. Start with Simple Evals-as-Code

Use a tool like Maxim to write test cases:

• Prompt: “Summarize this contract for a procurement manager.”

• Expected: “Must include payment terms, duration, and penalties.”

• Scoring logic: Pass/fail if coverage < 80%

This gets you repeatable, versioned, shareable tests.

3. Use Humans to Calibrate Early

Before you rely on GPT to grade GPT, use human raters to establish a ground truth.
You’ll learn:

• Where the model fails most often

• Which behaviors are easiest to miss

• What kinds of errors matter to end-users

Pro tip: Use 3–5 raters per sample and resolve disagreements. It’s time-consuming but builds trust in your scoring system.

4. Move to Automated + Continuous Evals

Once your rubrics stabilize, automate scoring:

• Use GPT-as-judge with labeled references

• Track evals in CI/CD (test before you deploy a new model)

• Monitor in production: hallucination rate, failure rate, drift

Connect feedback from real users to eval outcomes — and improve your model or retrieval strategy accordingly.

5. Maintain an Eval Leaderboard

Internally, track how different models, prompts, or configs perform over time:

• GPT-4-turbo vs Claude 3.5

• Base model vs fine-tuned

• Retrieval pipeline A vs B

This helps de-risk changes, justify investments, and communicate clearly with execs.

LLMs are entering critical infrastructure.

They’re advising doctors.
They’re writing policy.
They’re negotiating contracts.
They’re interacting with your customers.

Which means evals are not just technical hygiene. They are a core business capability.

In fact, as models become commoditized — evals become the moat.Your edge won’t come from the base model you use. It will come from how well you measure, tune, and align that model to your domain.

Big Tech will offer infinite memory next. But it’ll be a lock-in tactic. We need this layer to be open, sovereign, and user-owned.

I don’t want a hundred context engines.
I don’t want to be platform-loyal out of sunk-cost guilt.
I just want to stop going on first dates with my own data.

ChatGPT: Remembering Within Walls

OpenAI recently upgraded ChatGPT's memory capabilities, enabling it to "reference all of your past chats to provide more personalized responses."
It now learns your preferences, tone, and working style — creating a surprisingly personal experience within its ecosystem.

The critical limitation?This memory exists exclusively inside OpenAI’s walls.
Your carefully trained memory profile can’t travel with you. It’s trapped.

Claude: Facts Without Flow

Claude, Anthropic’s flagship AI assistant, made substantial memory strides with Claude 4.It can now extract and store key facts to build tacit knowledge over time.

The latest release, Claude Opus 4, introduced true memory persistence, solving what insiders called the “amnesia problem.” In one demo, it worked autonomously for seven hours on a complex coding project — showing unprecedented continuity.

Still, Claude struggles with cross-thread coherence.
It remembers facts, but not the flow.
Conversations live in silos, not across sessions.

Gemini: Google-Integrated Memory

Google’s Gemini is deeply integrated with Gmail, Calendar, Photos, and more — using a system internally known as “pcontext” (personalized context).

It can surface suggestions before you ask, tapping into your digital life.
It remembers your preferences, personal details, and work habits.

But almost entirely within the Googleverse.

Perplexity: Research Without Retention

Perplexity AI is exceptional at research. Its Deep Research mode performs dozens of searches and reasons across sources to deliver comprehensive answers.

But memory? Limited.

Currently, Perplexity operates with session-based memory.
It resets with every new conversation — losing context, continuity, and user history.

Beta memory features are in the works for late 2025.
These will begin auto-storing relevant information across threads — but as of now, you start from scratch every time.

The Core Problem: Siloed Memory

Across all these platforms, memory and personalization remain confined within proprietary ecosystems.

No AI system today offers true cross-platform memory transfer.
The result: users who’ve built deep profiles are stuck, facing significant switching costs just to try something new.

Why Platforms Resist Memory Portability

Let’s be honest: this isn’t a technical limitation.It’s a business strategy.

Memory is lock-in.The more context a platform captures, the harder it is for users to leave. This isn’t just a consumer problem. Enterprise leaders are already wary of AI vendor lock-in — just like they were with cloud and CRM platforms before.

Memory walls are the new moats.

The Repetition Tax

Every platform reset? I pay for it.

• Re-upload that doc.

• Re-contextualize my goals.

• Re-explain my writing tone.

Over. And over. Again. This is the repetition tax — and we’re all paying it. It wastes time, tokens, and energy.It turns powerful AI into glorified autocomplete. And it creates one of the biggest hidden costs in today’s AI economy.

The irony? These tools are supposed to save us time.

A Better Memory Stack: Beyond Human Recall

The right memory architecture isn’t a monolith. It’s a layered system:

• Basis Knowledge — foundational intelligence: timeless facts, core truths

• Logic Memory — rules and behaviors: what to do when

• Content Memory — reference and task material: past outputs, documents, structured context

When AI systems coordinate across these layers, they evolve from assistants to collaborators.

The First Platform That Gets Memory Right Will Reshape AI

Because here’s the truth:

The AI revolution won’t be won by whoever builds the smartest model.
It’ll be won by whoever builds the best memory — and gives you ownership of it.

The vault.
The protocol.
The plumbing.

It’s all missing — but I am it’s coming.A memory that travels with me, learns with me, and belongs to me. Because the future of AI isn’t just about intelligence.

It’s about remembrance — and control.

It does not happen all at once.
Just a quiet shift in how you see the work.

You begin to notice patterns.
You say "no" with clarity.
You anticipate what is coming.
And over time, you stop chasing outputs and start shaping outcomes.

Here are the signs that you are crossing that line:

1. You can predict behavior, not just ship features

You look at a new capability and ask,
“Will this actually change what users do?”

Not, “Will they try it?”
But how it will change their day, their workflow, or their decisions.

You start to notice the difference between adding functionality and creating real change.Your hit rate improves.
Not because of luck, but because your understanding of users has deepened.

2. You see weak strategies clearly—even if they look strong on the surface

A flashy competitor launch does not distract you.
You ask, “Why did they partner instead of build?”
“This looks impressive, but can they scale it?”
“Is this a feature, or is it a strategy?”

You stop reacting. You start diagnosing.
This clarity helps you focus your own energy where it matters most.

3. You are not building more. You are building what matters

You have learned the cost of trying to do everything.
You focus. You simplify. You commit.

Every roadmap decision ties back to a core business goal or user need.
You know that your team’s energy is limited.
You treat that energy as a resource to protect, not to spend freely.

4. You can shape adoption curves—not just hope for them

You know the difference between usage and need.
You have turned a capability that no one asked for into something they cannot imagine working without.

You have seen the flat line in the early weeks— and you have figured out what unblocks it.

You fix the activation gaps.
You tweak the UX where users stall.
You build pull—not just by features, but by habit loops, workflows, and urgency.

And when it clicks, it does not just grow. It spikes into a hockey stick

Not because of luck. Because you pushed the right levers.
Because you kept iterating long after the release.
And now, you can do it again—because you understand the anatomy of a hockey stick.

5. You can identify talent—often before others do

You do not just look at experience. You look for curiosity, ownership, and growth mindset.

You recognize when someone is ready to stretch. You give them space, not just tasks.
And over time, you build a team that scales with you.

6. You say “no” often—and the team gets better because of it

You used to say yes to keep momentum. Now you say no to keep direction.

You know that not every idea deserves space.
You protect the roadmap. You protect the team’s focus.
You protect the bet.

And because of that, the right work gets done better and faster.

7. You start thinking like a product marketer

You know every launch does not deserve a spotlight.
You decide: is this a blog, a press release, a webinar, or just a quiet update?

You position the product with intention. You choose moments that matter.
And you speak the language of the market—not just the roadmap.

8. You can quickly triage field asks without losing focus

The requests keep coming. Sales, success, partners, all asking for something.

You do not panic. You sort.
→ Shiny but low impact? Note it, move on.
→ Critical for long-term value? Prioritize.
→ Competitive gap? Fill it fast.

You are not reactive.
You are strategic.
That discipline keeps your plan intact.

There is no checklist that makes you a product leader.
No finish line.

But one day, you notice something:
You are no longer chasing wins.
You are building momentum.

And that changes everything.

An enterprise tool recently launched a glowing AI assistant—only for users to ignore it completely. Why? It floated awkwardly in the UI, offered generic responses, and solved nothing real. This isn't an outlier; it's a pattern.

While many companies are racing to slap AI labels on their features, most aren’t rethinking the product DNA beneath them. True transformation won’t come from generative widgets. It will come from re-architecting systems around intelligent behavior—systems that reason, adapt, and act without being prompted.

This isn’t just another tech cycle. It’s a full-stack product rethink.

From Features to Intelligent Systems

The diagram below illustrates the three-layer architecture that enables truly agentic AI systems:

1. User Interactions Layer — Where users engage with the system across multiple channels: Web, Mobile, Browser, Chat, and APIs.

2. Agentic AI Platform — The orchestration engine that coordinates workflows and agents to accomplish user goals.

3. Core Foundations — The bedrock layer of models, data systems, and tools that power intelligent behaviors.

This model represents a move away from bolted-on AI features toward deeply integrated systems that reason, act, and adapt autonomously.

The biggest failures I’ve seen aren’t because the AI didn’t “work.” It was because no one cared when it did.

The problem isn’t technical capability. It’s product relevance.

AI that doesn’t solve friction is just noise. And noisy features get muted, skipped, or disabled. Fast.

So before jumping into tools, agents, or architecture—let’s ground ourselves in five principles that separate AI that gets adopted from AI that gets abandoned.

Principle 1: Design for Clarity, Not Cleverness

What Fails: A generic chatbot that lives in your CRM and offers to "help with anything." No context. No value. No adoption.

What Works: Context-aware assistants that offer value exactly when needed:

• "This customer matches 3 recent wins — want to review the decision drivers?"

• "This objection pattern has appeared in 18% of lost deals — here’s how top reps handle it."

The best AI experiences feel inevitable, not optional. They blend seamlessly into workflows instead of demanding new behavior.

Principle 2: Deliver Measurable Value Before the Next Budget Cycle

In the enterprise world, patience is expensive. Leaders don’t buy AI—they buy time back, improved forecast confidence, or the ability to close more deals. If your AI feature can’t prove value before the next QBR, you’ve already lost the renewal battle.

Think less in terms of product milestones and more in terms of business velocity accelerators:

Don't bury impact in quarterly decks. Ship it into the workflow. Let users feel the win—“AI saved you 3.2 hours this week.”

Speed compounds. Every day you ship measurable value is a day you build trust.

Principle 3: Adoption Requires Behavior Change

Launch ≠ Adoption. Real adoption means changing how people work—and that takes thoughtful design.

Most teams underestimate how much behavior inertia they're up against. New tools—even magical ones—get ignored if they’re not intuitive, habitual, and clearly valuable. You can’t brute-force change; you have to embed it.

What works:

• Meet users in their flow – Think Slack commands, email plugins, browser nudges. Don’t pull users into your product—bring intelligence into theirs.

• Make benefits undeniable – Instead of saying, “AI helped,” show: “This insight saved your team 6.5 hours last week.”

• Trigger habit loops – Small, daily wins build trust. Celebrate those wins with nudges, dashboards, or peer comparisons.

And remember, adoption ≠ universal usage. Adoption means the right people, using the right AI, at the right moments. That often means both copilots and agents coexist.

Copilots shine when the user is in control and needs augmentation: writing content, debugging code, prepping for meetings.

Agents shine when the system should take the lead: automating renewals, monitoring risk, flagging deal anomalies.

The job isn’t to force one model—it’s to define the right model per moment of intent.

Adoption design is behavior design. And behavior is shaped by clarity, reward, and repetition.

Principle 4: The Real Shift is from Copilots to Agents

Copilots are helpful. But agents are transformational.

Copilots wait for commands. Agents chase outcomes.

This distinction changes how you architect, how you scale, and how your users trust the system.

Aspect Copilot Agent Activation User-initiated Self-initiated Scope Single tasks Multi-step workflows Learning Static Improves over time Autonomy Suggests Acts

Real Agent Scenarios:

• A revenue ops agent notices conversion dips, attributes it to a competitor’s pricing shift, and launches a counterplay suggestion to sales leaders—no human prompt required.

• A customer health agent preempts churn by pulling usage signals, assigning a targeted reactivation campaign, and surfacing only the top-3 at-risk accounts to human CS reps.

Agents don't just assist—they operate. Which means you need:

• Goal-based instruction engines

• Stateful memory

• Feedback-boundaries and escalation protocols

The shift to agents is not about smarter responses. It's about smarter decisions without a prompt.

Principle 5: Tooling Shapes Execution Velocity

The speed at which your team learns—and ships—is now a competitive advantage.

Modern tooling isn’t about convenience. It’s about compressing cycles of alignment, experimentation, and feedback.

Bolt helps teams:

• Replace static PRDs with living workflows

• Drive constant alignment through linked context maps

• Shift planning from “deadlines” to “decision checkpoints”

Cursor helps:

• Engineering and product pair in real time with LLMs

• Speed up code reviews, POCs, and iteration loops

Real example: At one org, a team using Bolt cut feature rollout cycles from 6 weeks to 10 days—because context didn’t get lost in translation. Cursor allowed engineers to test API ideas live with the PM instead of waiting 2–3 days for design grooming.

The result? A faster path to validated learning. Fewer handoffs. More outcome ownership.

The real unlock isn’t that AI is helping build the product. It’s that it’s helping teams become more agentic themselves—less reactive, more autonomous, and relentlessly learning.

In an AI-first world, your execution stack matters as much as your model stack.

Agentic AI Architecture

Most teams still think in terms of layers: UI → Backend → Data. But in the world of agentic systems, those lines blur. What we’re really building is a loop of perception, decision, and action — one that needs context, trust, and autonomy at every step.

Let’s walk through the architecture:

1. Core Foundations (Data, Models, Skills, Tools) This is the base. Think of it as your AI system’s nutrition and wiring.

   If your data is messy, your AI outputs will be too. If your models can’t adapt, they’ll quickly fall behind. And if your developers and agents lack the right tools and skills, you’ll stall before you start.

   You need:

   • Clean, reusable data pipelines

   • AI models you can tune and trust

   • Agent tools like retrievers, planners, scorers, and memory handlers

   • Skills that include prompt engineering, system design, and reinforcement strategy tuning

   This isn’t just infrastructure—it’s capability. Without this foundation, everything above it crumbles.

2. Agentic AI Platform (Orchestration, Reasoning, Action)
   This is the living brain. It handles:

   • Goal-directed planning (“Given this input, what should I do?”)

   • Context switching and state retention

   • Feedback integration: “Did the action succeed? Should I try again differently?”

   Think of it as a dynamic operating system, not a single algorithm.

3. User Interactions (Web, Mobile, API, Chat)
   This isn’t just UI—it’s trust-building surface area. Each interaction must:

   • Make the AI feel predictable and supportive

   • Show value before asking for attention

   • Offer transparency and graceful exits when needed

When done well, this architecture creates a virtuous cycle: The more the agent helps, the more users rely on it. The more it’s used, the smarter and more aligned it becomes.

This is what separates systems that delight from those that confuse. The best AI doesn’t ask to be used — it becomes part of how work gets done.

Your 60-Day Roadmap

• Weeks 1–2: Map workflows. Identify highest-friction points.

• Weeks 3–4: Prototype AI nudges or task automations.

• Weeks 5–6: Launch with measurement tied to specific KPIs.

• Weeks 7–8: Refine. Introduce memory. Explore agent automation.

Final Word

The winners in the AI era won’t be those with the most advanced models. They’ll be the ones who integrate intelligence invisibly into the core of their product.

Start with user friction. Where are your users stuck, slow, or improvising? That’s your roadmap.

Map pain before AI — the model should come last.
Prove value fast — days, not quarters.
Reinforce new behavior — not just surface delight.
Architect systems — not just interfaces.

Agentic AI isn’t about smarter assistants. It’s about building intelligent infrastructure that works for your users — even when they’re not watching.

And here’s the risk: if you don’t build this, someone else will—faster, quieter, and with better compounding returns. The cost of delay isn’t just missed innovation—it’s irrelevance. AI isn’t a sidecar anymore. It’s the chassis.

You’re not deciding whether to build agentic systems. You’re deciding whether your product deserves to survive the next evolution of user expectations.

And yet, industries evolve. Markets shift. What worked yesterday might be completely useless tomorrow. So why do we keep playing by the same old rules?

The answer: Most people don’t stop to question them.

Enter first principles thinking—a mental model that helps you break free from conventional wisdom and build solutions from the ground up. Not by tweaking what exists, but by stripping everything down to its core truths and starting fresh.

A "first principle" is a foundational assumption or proposition - it is foundational in that it cannot be deduced from other assumptions or propositions.

Think of a first principle like an element. It cannot be broken down further. It is pure.

It’s how Elon Musk redefined space travel. How Apple changed how we interact with technology. How Stripe made payments seamless.

And if you’re in product, strategy, or business, it’s how you stop getting stuck in incremental improvements and start driving real innovation.

What is First Principles Thinking, Really?

First principles thinking is problem-solving at its purest form. It forces you to remove assumptions, question legacy ideas, and rebuild based on foundational truths.

It’s different from reasoning by analogy, where you assume things must be a certain way because they’ve always been done that way.

Instead, it’s about asking:

• What do we know to be absolutely true?

• What assumptions are we making that might be wrong?

• If we started from scratch, how would we solve this?

Farnam Street explains it like this:

"Reasoning by first principles removes the impurity of assumptions and conventions. It’s a way of thinking that allows you to generate original solutions rather than simply following paths that already exist."

This is why first principles thinking is so powerful in product management, business strategy, and innovation.

Step 1: Deconstruct the Problem

The first step in first principles thinking is identifying absolute truths—things that don’t change no matter what industry you’re in.

Some examples:

• Buyers don’t care about your product. They care about their problems.

• Knowledge doesn’t equal skill. Just because someone knows something doesn’t mean they can do it.

• More features don’t mean a better product. A cluttered product is just a graveyard of ideas.

These are truths. They don’t change. They don’t rely on trends. They are the foundation upon which everything else should be built.

Yet, most companies ignore them. Instead, they make incremental improvements to broken systems.

Step 2: Challenge Assumptions

Once you’ve established the fundamental truths, the next step is to question the assumptions that have built up around them.

For example:

• Making rockets is expensive. Is it? Or is that just how the aerospace industry operates?

• More training leads to better performance. Or does it just lead to more content consumption?

• More money means better outcomes. Or do inefficient processes eat up most of that investment?

These assumptions often come from years—sometimes decades—of doing things a certain way. But that doesn’t mean they’re true.

First principles thinking forces us to ask: What if we started from scratch?

Step 3: Rebuild from Scratch

This is where real innovation happens. Instead of iterating on broken models, you build new ones from the fundamental truths you’ve identified.

Examples:

• SpaceX: Everyone thought rockets had to be insanely expensive. Musk broke it down—rockets are made of aluminum, titanium, and carbon fiber. The raw materials cost far less than the final product. The real expense was inefficient supply chains and manufacturing. Solution? Build rockets in-house, reuse them, and slash costs.

• Stripe: Payments were slow, painful, and bureaucratic. The assumption? That businesses had to go through banks and clunky legacy systems. Stripe asked: “Why?” and built a seamless, developer-first API that changed the game.

• Apple: Before 2007, phones had physical keyboards. Everyone assumed that’s just how phones worked. Apple stripped it down to first principles. What’s a phone for? Communication, access to the internet, apps. Do you need buttons for that? No. The iPhone was born.

These aren’t just product innovations. They’re mental shifts that redefined industries.

What Does a First Principles Product Strategy Look Like?

Most product roadmaps look like this:

• Competitor X is doing this, so we should too.

• Customers asked for this feature, let’s add it.

• Leadership wants this because it’s part of the strategy deck.

None of these are first principles approaches. They’re reactive. They stack ideas on top of existing ones instead of questioning if the foundation is even solid.

A first principles approach starts differently:

• What is the user actually trying to do?

• What frustrates them the most?

• What workarounds do they use?

For example, consider sales reps. A product team might assume, “Sales reps need better training materials.”

But a first principles approach would ask:

• Do sellers really want to train more? No. They want to win more deals.

• Do sellers want to be top of their class? No. They want to show off the deal they just closed.

• Do sellers want content? Not really. They want real-world tactics that help them win.

The Mindset Shift

First principles thinking isn’t just about better products. It’s about breaking out of old habits and seeing problems for what they really are.

It forces you to:

• Ignore industry norms because they’re often based on outdated assumptions.

• Stop looking sideways at competitors because copying them means you’ll always be one step behind.

• Start solving problems from the ground up because that’s how you build something truly different.

So, What’s the One Truth in Your Industry That’s Actually an Outdated Assumption?

Every industry is built on assumptions that were true once—but may not be true anymore.

What if you challenged them?

Because playing by old rules is how you get old results. Rewriting the rules? That’s where real innovation happens.

With a last-minute booking based on a 4.5+ tripadvisor rating, I expecting a nice, modern , comfortable and central place to crash after long days of exploring - nothing fancy or luxury.

But within minutes of stepping inside, I realized this place had something very special!

It was different.

• The fragrance that greeted us at the entrance— was a blend of fresh citrus, white tea —clean, uplifting, and luxurious, similar to global luxury chains like Westin (white tea and aloe) and The Ritz-Carlton (signature cedar and citrus blends).

• The bright yellow and white branding, simple yet striking.

• The community-like vibe—a mix of travellers, business folks, and digital nomads working on their laptops in the common area. Cookies, beverages and other freebies in the passage ways

• The tech-enabled room experience—check-in via app, room service at the tap of a button, and keyless entry that actually worked.

• The thoughtful little touches—a doodle-style city map highlighting must-see places with estimated travel times, so you could plan your day in seconds.

And then there was the location—6 minutes from Connaught Place, India Gate, Jantar Mantar, and Gurudwara Bangla Sahib. Every key landmark was just around the corner.

By the time I checked out, my mind wasn’t just thinking about the hotel experience. It was thinking business.

I knew I had to dig deeper into the numbers behind Bloomrooms.

The Growth Story: From ₹49 Cr to ₹250 Cr in 3 Years 📈

Bloomrooms isn’t just a cool brand with great interiors. It’s a hyper-efficient, asset-light, tech-driven business that’s scaling fast.

Here’s how the numbers stack up:

• FY22 Revenue: ₹49 Cr

• FY23 Revenue: ₹144 Cr (3x growth in one year)

• FY24 Revenue: ₹250 Cr (+73.6% YoY growth)

• FY24 Profit: ₹14 Cr (2.3x increase from FY23)

That’s not just growth. That’s momentum.

Bloomrooms has gone from a niche player to a fast-scaling mid-market disruptor in India’s hotel industry. And this is just the beginning.

Here’s Why I think Bloomrooms is Poised for Massive Success 🚀

1. The Asset-Light Model: Scaling Without Owning

Bloomrooms doesn’t own most of its properties. Instead, it leases, franchises, or manages hotels under its brand.

This means:

• Less capital investment

• Faster expansion

• Higher profitability per hotel

It’s the same playbook that helped Marriott and OYO scale globally—grow without the baggage of real estate ownership.

Right now, Bloomrooms operates 50+ hotels across India and is on track to hit 100+ locations in the next 2-3 years. And unlike many hospitality startups that burn cash, they’re laser-focused on profitable growth.

2. Premium Budget Positioning = Higher Margins

Unlike OYO or FabHotels, which target the absolute low-cost segment, Bloomrooms sits in the premium budget category—the sweet spot between affordability and comfort.

Their target audience?

• Business travelers who want efficiency without luxury pricing.

• Digital nomads and young professionals.

• Family travelers looking for clean, centrally-located hotels.

This positioning allows them to charge higher-than-budget rates ( ₹4,000 - ₹ 9000 per night) while maintaining lean operations. That’s a recipe for high margin expansion.

3. Tech-Driven Efficiency = Lower Costs, Better Experience

One of the things that impressed me most at Bloomrooms @ Janpath was how seamless the experience felt.

• Mobile check-in & check-out—No front desk hassle.

• App-based room service & support—Ordered extra towels in seconds.

• Keyless entry—No misplaced card drama.

• Automated room pricing—AI-driven dynamic rates, like airlines.

I did not have to call room service for anything , just an app chat was getting things done, I see how GenZ and Gen Alpha would totally love this

All of this means lower operational costs, faster turnaround, and higher guest satisfaction.

4. "Bloomverse" Branding: Every Detail Counts

Everything at Bloomrooms is branded to perfection. The bright yellow theme runs through the entire stay experience:

• Yellow accents in the shower, lamp shades, and furniture.

• A local city map in yellow that highlights must-see spots and distances.

• Even the key card and toiletries stay true to the brand aesthetic.

  

It’s not just a stay; it’s an experience. Welcome to the Bloomverse.

5. Common Spaces = A Hostel-Like Community Feel

Unlike traditional hotels that keep people locked in their rooms, Bloom encourages interaction.

• Lounge spaces with coffee, workstations, and books.

• Free tea and coffee in common areas.

• A casual, open seating vibe that feels like a premium hostel.

This hybrid model—a mix of business hotel + premium hostel vibes—is perfect for modern travelers who want both privacy and social engagement.

6. Location, Location, Location!

The biggest advantage? You save time.

At ₹9,000 per night, we got a suite room in the heart of Janpath—two separate rooms, pod-style interiors, and all the amenities of a 4-star hotel inside the room. And the location? Absolutely unbeatable:

• Connaught Place - 5 mins

• India Gate - 6 mins

• Red Fort - 20 mins

• Rashtrapati Bhavan - 6 mins

• Jantar Mantar - 5 mins

• Supreme Court - 8 mins

• Lok Sabha - 7 mins

• Khan Market - 12 mins

Being this central meant we could do so much more in less time. All other options in this area were at least ₹17,000++ for a family of 3.

7. Direct Booking Strategy = Higher Profits

Most hotels rely on OTAs like Booking.com and MakeMyTrip, which take a 15-25% cut. Bloom is aggressively driving direct bookings via its website and app, reducing dependency on third-party platforms.

Direct bookings = higher margins + better customer retention.

8. Localized Experiences in Iconic Cities

The other thing I found, as I dug deeper is that Bloomrooms is not just about uniformity—it’s about bringing local character into the guest experience while maintaining its signature service. The brand has expanded to multiple cities, each offering a unique stay:

• Mumbai (Worli, Juhu, Andheri) – Business-focused properties in the heart of the city.

• Udaipur (The Bloom Kothi) – A beautifully restored heritage-style property that blends modern amenities with the charm of a local haveli.

  

• Goa (Candolim, Calangute) – A boutique coastal experience infused with laid-back beach vibes.

  

• Bangalore (Indiranagar, Brookefield) – Perfect for business and leisure travelers alike.

• Pune (Koregaon Park) – A mix of work and nightlife

By localizing the experience—like the kothi-style stay in Udaipur—Bloomrooms ensures that guests don’t just visit a city; they experience it authentically while enjoying the reliability of a branded stay.

The Big Picture: What’s Next for Bloom Hotels?

Bloom Hotels is already profitable and is now looking to raise ₹400 Cr to fuel expansion.

If they execute well, Bloom Hotels could be the next big hospitality success story in India.

They’ve already grown 5x in three years. And I’d bet they’re just getting started.
And personally for me, couldn’t help but keep hum this all through the stay :)

Look at the stars

Look how they shine for you

And everything you do

Yeah, they were all yellow

I’d stay again in a heartbeat. No second thoughts, just a blink, and I’m booking!

I feel this shift firsthand every day. I’m constantly testing out new AI tools, spinning up prototypes with Bolt to see what’s actually possible, and experiencing the latest AI-powered products on Product Hunt. I read everything I can—technical blogs, research papers, deep dives from builders who are pushing boundaries. Because it’s not just about keeping up; it’s about developing an instinct for where things are headed before they arrive.

Some will lead. Some will follow. The difference comes down to how well you adapt to an environment where the ground shifts every 60 days.

Seeing Around the Corner

AI doesn’t evolve gradually—it leaps forward in unpredictable bursts. A roadmap built six months ago is already outdated if it didn’t anticipate the latest breakthroughs. The only way to stay ahead is to build optionality into your plans. Shorter cycles. Running two roadmaps at once—one for what's possible today, another for what AI will make possible soon.

If AI can already do what you planned to build, your roadmap needs a reset. If you’re waiting for things to stabilize, you’re already behind.

AI Gimmicks vs. Real Customer Value

The fastest way to burn engineering time? Shipping AI features that look impressive but don’t solve a real problem. AI-generated content, chatbots, predictive analytics—none of it matters unless it drives revenue, retention, or operational efficiency. A simple test: if you removed the AI label, would customers still pay for it? If not, it’s a gimmick.

AI should make workflows disappear, not create new ones. It should automate real pain points, reduce friction, and help customers make better decisions, faster. If your AI feature requires a user manual, you’re doing it wrong. A lesson learnt hard way - the expectation to change user behavior now that you have AI is a huge mistake, don’t make things hard for them to discover , ensuring intuition led experiences is critical

The New Reality of Product Roadmaps

AI doesn’t just change what gets built—it changes how product teams operate. The old model of stable problem spaces and long-term planning no longer works. What seemed impossible to automate last year is now trivial. What was cutting-edge six months ago is now a commodity.

The best product teams don’t just plan for AI advances; they expect them. Shorter planning cycles. Multiple tracks running in parallel—one for what’s possible today, another for what’s likely to emerge soon. A ruthless willingness to kill projects that don’t move the needle because AI has already made them obsolete. The roadmap isn’t a fixed path—it’s a system that evolves as fast as AI itself.

First Principles Thinking Over AI Hype

AI can hallucinate, misfire, and introduce unpredictability. That’s why product leaders can’t rely on models to do their thinking for them. The foundation must be first principles. What problem are we solving? What’s the simplest, most reliable way to deliver value? AI is a means to an end, not the starting point.

Just because AI can do something doesn’t mean it should. Do users actually want full automation, or do they need a better assistive workflow? Does AI improve decision-making, or does it introduce unnecessary complexity? The best product teams ask these questions before a single line of code is written.

AI Isn’t a Feature—It’s an Organizational Capability

AI is not a product add-on. It’s an entire capability that needs to be built into how a company operates. That means data pipelines that aren’t a mess, feedback loops that continuously improve models, and engineering teams that understand how to balance AI complexity with usability.

It also means understanding AI’s cost structure. Every AI-powered feature has a real-time operational cost. Compute, model training, API calls—it all adds up. If your AI isn’t delivering more value than it costs to run, it’s a liability. The smartest companies bake AI economics into product decisions early, not after launch.

Trust and AI UX

AI adoption isn’t just about functionality—it’s about trust. Users don’t care about model architectures or neural networks. They care about whether AI helps them without failing unpredictably. One hallucination, one bad decision, one broken experience—and trust evaporates.

That’s why AI UX must be built around predictability, control, and explainability. Show users why AI made a decision. Give them the ability to override it. Make it assistive, not intrusive. AI that confuses or frustrates users isn’t a feature—it’s a risk.

Monetizing AI Without Giving It Away

AI shifts the way software is priced. Unlike traditional SaaS, AI-powered features consume resources every time they’re used. That changes the economics of pricing. Some companies will bundle AI into existing plans, others will charge for premium capabilities, and some will move toward usage-based pricing, and some towards Outcome based.I’ve been tracking companies that have successfully charged on outcomes, not just usage—because that’s where the real AI pricing frontier is.

If AI helps customers sell more, close deals faster, or automate work, there’s a price tag for that.

Measuring AI Success Beyond Adoption

Traditional product metrics don’t work for AI. Adoption alone doesn’t tell you if an AI feature is useful. Instead, track real impact. How often do users correct AI decisions? Is it saving time or adding friction? Are outcomes improving?

A good AI feature isn’t just used—it’s trusted. If you’re not tracking how often users override AI suggestions, how much time it actually saves, or whether it drives better decisions, you’re flying blind. AI is never finished. If you’re not constantly iterating, it’s getting worse.

The AI race is unforgiving. Every breakthrough compresses competitive timelines. Companies that integrate AI quickly will dominate. Companies that hesitate will struggle to catch up. Speed matters more than ever—not just speed of execution, but speed of learning, adapting, and iterating.

The best AI-first companies don’t just react to change; they force it. They run constant experiments. They launch before the tech is perfect, then refine in real-time. They create new AI-driven moats before competitors even see the opportunity. AI isn’t just a battleground, it’s an arms race.

Good product leadership has always meant anticipating trends, separating hype from real value, and rallying teams around the right bets. But AI changes the stakes. The ones who get it right will define the next generation of SaaS. The ones who don’t will spend the next decade reacting (if they get to stay around)

Disclaimer - I’m an active Instagram user. I spend about 20 minutes daily scrolling, liking, and occasionally posting. But on a recent trip to Udaipur, I came face-to-face with a reality that left me unsettled.

The hotel I stayed at had gone viral on Instagram a few months back, attracting travelers from all corners of the country and outside too. Its restaurant, open to the general public, had become something of a hotspot. Over the course of four nights, I observed a pattern that became impossible to ignore.

The hotel had a lake-facing restaurant open to people not staying in the property too. Every group coming in for a meal spent an average of two hours on the property. Of that time, more than half was spent making reels. Not eating, not talking, not soaking in the beautiful surroundings—just creating content.

It was surreal. People weren’t visiting to experience the place—they were there to capture it. The food was a backdrop. The company, secondary. Their focus was on curating a moment for their followers rather than enjoying one for themselves.

It made me pause and reflect, and write this up.

This post is a little different from what I usually write. But today, I want to step back and examine a product not as a professional, but as a person. A user. A parent.

Instagram is a fascinating product. It’s beautifully designed, endlessly engaging, and undeniably successful. But when we zoom out, when we look at the human experience, I think there’s a conversation worth having about the role it plays in our lives—and the cost it imposes on our most valuable resource: time.

Instagram is product designed to connect us, inspire us, and empower us to share our lives. But beneath its glossy veneer lies a sobering truth: it’s quietly siphoning away something irreplaceable—our time.

When you’re scrolling through reels or stories, you don’t notice it. It’s just a few minutes, maybe a quick dopamine hit. But zoom out.Multiply that “just a few minutes” by millions of people around the world. Every hour, Instagram users collectively spend billions of minutes scrolling through a feed designed to keep them hooked. In a single day, we’re talking about trillions of minutes—a staggering amount of human potential consumed, second by second, for something that won’t leave a lasting impact. That’s such a great cumulative loss of productive time! The magnitude is insane

Think about it. In ten years, you won’t remember the perfectly framed sunset you saw on someone’s profile. You won’t recall the quirky reel someone spent hours editing to sync perfectly with trending audio. These moments won’t matter, not because they were inherently bad, but because they were impermanent. They weren’t built to last.

What’s worse, many of these moments don’t even evoke genuine emotion. They might make you laugh briefly or give you a fleeting sense of envy or awe.

Contrast that with the things that do stick with you—the confidence you felt post something you mastered, the time your best friend made you laugh so hard you cried, the feeling of awe as you watched a real sunset with your own eyes, or the quiet comfort of a shared meal with someone you love. These moments don’t just make you feel something in the moment; they become part of you. They shape how you see the world, how you remember your life.

The sad truth is that Instagram replaces these real, enduring experiences with transient, curated glimpses of other people’s lives. It robs us of our presence, swapping authentic connection for shallow consumption. And in doing so, it wastes the one thing we can never get back—our time.

This is the hidden cost of Instagram. It’s not the minutes spent scrolling; it’s the moments we miss while we’re busy doing it. Moments that could have been spent laughing, creating, learning, or simply living. Moments we’ll never get to relive.

So the question we must ask ourselves is this: Are we spending our time on things that truly matter? Or are we letting it slip away, one scroll at a time?

This isn’t a critique of Instagram’s design or its team’s intent. It’s a critique of a system—a system that incentivizes engagement at all costs and measures success in minutes spent rather than value delivered.

Time is our most precious resource. It’s finite, non-renewable, and utterly priceless. Yet, products like Instagram have found a way to monetize it, one scroll at a time.

Think about it: how often do you hear someone say, “I’ll just check Instagram for a minute,” only to lose an hour? This isn’t accidental. It’s the result of carefully crafted product decisions—endless feeds, autoplay videos, and those addictive little red notification dots.

These aren’t just features. They’re mechanisms. Mechanisms designed to keep you engaged, regardless of whether that engagement is meaningful.

But here’s the thing: meaningful engagement can’t be measured in minutes. It’s measured in impact. Did this product help me grow? Did it strengthen my relationships? Did it leave me better than it found me?

For many, the answer is no. Instead, it leaves us drained, distracted, and disconnected from the people and experiences that truly matter.

And that brings me to a very controversial thought - Not every product needed to be made, and maybe Instagram was one of them.

That’s a bold claim, I know. But hear me out. At its core, Instagram isn’t solving a real human problem. It’s exploiting one: our need for validation, our fear of missing out, and our desire for connection in an increasingly disconnected world.

Instead of fulfilling those needs, it amplifies them. It creates a world where self-worth is measured in likes, where memories are curated for an audience rather than cherished for their own sake, and where time is spent performing rather than living.

There’s a beautiful concept in a book I recently read, and highly recommend too Same as Ever by Morgan Housel: there are two types of information: permanent and expiring. Permanent information, found in books or deep reflections, compounds over time. It teaches us not just what happened, but why it happened—offering insights that remain relevant for years or even decades.

Expiring information, on the other hand, is fleeting. It tells us what happened now but holds little long-term value. Instagram is a masterclass in expiring information. It keeps our short attention spans occupied and hooks us with the urgency to consume before the content loses relevance.

But what do we gain from this endless chase? In 10 years, will you remember the reel you laughed at on a random Tuesday? Probably not. But you might remember a quiet conversation with a loved one, a book that shifted your perspective, or an idea that stayed with you long after you encountered it.

By filling our days with expiring information, we crowd out the space for reflection, growth, and connection—the things that truly matter.

As product leaders, we have a choice. We can design products that extract value from people, or we can design products that add value to their lives.

The latter is harder. It requires us to resist the allure of engagement metrics and focus instead on outcomes that are harder to measure but infinitely more meaningful.

Some principles to guide us product people, therefore

1. Time is Sacred.
   Every feature, interaction, or decision should respect the user’s time. Instead of asking, “How can we make people spend more time on this product?” ask, “How can we help them achieve their goals and get back to their lives?”

2. Design for Impact, Not Engagement.
   Engagement is a vanity metric. The true impact is about enabling growth, fostering connection, or solving real problems. Measure what matters.

3. Optimize for Long-Term Value.
   Products should leave people better than they found them. That means thinking beyond the immediate interaction and considering the long-term effects on users’ well-being, relationships, and potential.

Every now and then, you come across a book that doesn’t just explain concepts but fundamentally shifts how you think about innovation. Innovation: The Five Disciplines for Creating What Customers Want is one such book. It breaks innovation down into clear, actionable principles and connects it all back to the ultimate goal—delivering customer value.

What stood out to me was how practical and relatable the insights are. From identifying unmet needs to building high-performing teams, the book offers a blueprint for anyone serious about creating products that customers truly want.

I found it quite impactful that so decided to distill the key ideas into this summary—both as a reference for myself and as something to share with others. If you're looking to drive meaningful innovation, this book is worth the read. Here's what I learned.

Every company wants to innovate.
Few actually do.

Why?
They stumble in the gaps.

Most teams don’t fail because they lack good ideas.
They fail because they don’t execute them well.

You’ve seen it happen:

• A flashy product launch fades into obscurity.

• A rebrand excites employees but falls flat with customers.

• A task force starts with enthusiasm but quietly fizzles out.

Here’s the hard truth: Innovation isn’t a one-time event. It’s a cycle.
When companies fail to see this, they resort to surface-level fixes:
Symbolic changes. Band-aid solutions. New paint on old problems.

If you want a culture of lasting innovation, you need to understand the cycles, the traps, and—most importantly—the steps to break through.

The Life Cycle of Every Product or Service

Every idea follows three stages:

Stage A: The New Concept
Excitement runs high. The idea is fresh. Teams pour resources into making it real.

Stage B: The Growth Stage
The product enters the market. Profits emerge. Teams ride the wave of success.

Stage C: Maturity and Decline
Growth plateaus. Value stagnates. Costs creep up. Without change, decline begins.

This is where most organizations stumble. Instead of driving real change, they fall back on symbolic fixes:

• Rebranding a product that still doesn’t solve customer problems.

• Launching “innovation task forces” that discuss but never execute.

• Talking only to supporters, ignoring skeptics.

The result? Decline happens quietly.

Understand How Customers Measure Value

Value isn’t what you say it is. It’s what the customer decides.
And customers are ruthless mathematicians:

Customer Value = Benefits - Costs

Take Herman, an environmentally conscious car buyer. He compares the Toyota Prius and the Ford Taurus:

For eco-conscious buyers, the Prius delivered clear value:

• Benefits: Pollution reduction, fuel efficiency, and a green identity.

• Costs: Slightly higher upfront price, but lower gas costs better tax incentives

The Taurus, while reliable, didn’t match what these customers valued most.

Here’s the breakdown:

What Drives Real Innovation

Innovation is the process of moving from Stage A to Stage B—creating and delivering new customer value in the market. This value must also enable sufficient profit for company growth.

The conceptual vehicle for getting there is your value proposition, which answers four critical questions:

1. What is the important unmet customer and market need?

2. What is your approach to address this need?

3. What are the benefits per costs of your approach?

4. Why are your benefits per costs superior to the competition?

Case Study: The Evolution of Video-on-Demand (VoD)

What was the unmet need?

Customers:

• Hated trips to rental stores.

• Resented late fees for overdue returns.

Market Gap:

• No seamless way to access movies. Customers craved:

  • Instant access to entertainment.

  • Freedom from late fees.

  • Convenience without store visits.

What was the approach?

VoD Systems:

• Instant access: Watch movies directly on TVs via cable or broadband.

• No late fees: Eliminated physical returns entirely.

• Subscription model: Replaced unpredictable costs with flat monthly pricing.

Key Innovations:

• Leveraged existing cable infrastructure, reducing costs.

• Created a “watch whenever” experience.

What were the benefits per costs?

Benefits:

• Convenience: No trips to stores, no deadlines.

• Cost Savings: Flat subscription fees replaced punitive late fees.

• Flexibility: Access to extensive libraries without stock limits.

Costs:

• Used existing cable/broadband infrastructure, avoiding expensive deployments.

• Offered low subscription costs relative to the value delivered.

Why was it superior to competitors?

Compared to Blockbuster:

• Convenience: No trips, no returns, no late fees.

• Cost: Transparent pricing over unpredictable penalties.

• Flexibility: Instant access to larger libraries, unrestricted by physical stock.

VoD didn’t just improve on traditional models; it redefined customer expectations, setting a new standard for convenience and value.

Another example Airbnb:

Unmet Need:

• Travelers wanted affordable, authentic alternatives to hotels.

• Hosts wanted a way to monetize spare rooms.

Approach:

• An online marketplace connecting travelers and hosts.

Benefits vs. Costs:

• Benefits: Affordable stays, unique accommodations, local experiences.

• Costs: Trust concerns, unfamiliarity with the model.

Why Airbnb Won:
Their benefits were superior to hotels: lower costs, better experiences, and unmatched scalability.

Why Innovation Fails

If innovation were easy, every company would nail it.
But many fall into the same traps:

1. Grand Visions Without Groundwork:
   Leaders announce bold plans but fail to involve teams in execution.

2. Symbolic Changes:
   Flashy rebrands or PR campaigns that don’t solve customer pain points.

3. Ignoring Resistance:
   Skeptics are dismissed as negative instead of being engaged as problem-solvers.

Innovation isn’t a Big Bang. It’s a system.

Fixing Your Innovation System

Here’s how to break the cycle and build lasting momentum:

1. Understand Your Customer Value Equation:
   Focus on what customers truly care about.

2. Spot White-Space Opportunities:
   Identify gaps where your product can win.

3. Avoid the Failure Formula:
   Include skeptics in the process.

4. Eliminate Muda (Waste):
   Cut unnecessary meetings and processes that don’t solve real problems.

5. Show Progress:
   Build trust through small, visible wins.

6. Remove Barriers:
   Don’t be afraid to start fresh when needed.

7. Spin the Flywheel:
   Incremental, steady progress beats flashy fixes every time.

High-performing teams are the engine of innovation.

Here’s what makes them work:

• A shared, crystal-clear vision that aligns everyone.

• Complementary roles where each person brings unique value.

• Collective rewards that recognize team contributions, not just individual effort.

• Respectful, open communication that fosters collaboration.

And most importantly:

They focus relentlessly on actionable outcomes. No wasted effort. No empty motions. Just meaningful progress.

The Bottom Line

Innovation doesn’t fail because teams don’t care.
It fails because they get stuck in patterns that don’t work.

Break the patterns. Build momentum.And start solving problems that matter.

Not all skills are created equal—some are visible, easily recognized and measured, while others are invisible, subtle, and harder to quantify. Visible skills get you noticed. Invisible skills get the job done. This distinction is crucial for understanding why certain leaders excel in specific environments and roles.

The Paradox of Product Leadership

Success in product leadership isn't just about what you see. It's about what you don't.

Consider two leaders: Cassie and Jessica. Cassie has impeccable visible skills—strategic thinking, flawless presentations, and sharp market research. Jessica, on the other hand, excels in invisible skills—system thinking, first principles thinking, and emotional intelligence. Who's more successful? The answer isn't as straightforward as you might think.

Visible Skills: The Obvious Hallmarks of Leadership

Let's start with Cassie. Cassie is a star in meetings. Strategic thinking? Check. . Market research? Check

Cassie's presentations are pixel-perfect, delivering clear, concise communication to stakeholders.

Leadership and stakeholder management? Cassie leads cross-functional teams and navigates complex relationships with multiple stakeholders, balancing differing priorities and expectations. Impeccable control of numbers? Absolutely. Cassie knows key KPIs and metrics inside out.

However, Cassie lacks a fundamental understanding of the persona, their unmet needs, and their jobs to be done. She’s great with numbers and strategy, but she doesn’t connect deeply with the customer’s experience.

Now, let’s look at Jessica. Jessica’s strengths aren’t immediately obvious. System thinking allows Jessica to understand the product within the larger ecosystem, seeing interdependencies and the ripple effects of every change. First principles thinking helps Jessica break down problems to their fundamental truths and build solutions from the ground up.

Emotional intelligence? Jessica excels at managing emotions—both personal and others'. Jessica builds strong relationships, handles conflicts effectively, and fosters a positive team environment. Adaptability? Jessica is flexible and open to change, pivoting quickly when new information arises. Resilience allows Jessica to handle stress and setbacks with a positive attitude, maintaining motivation and focus despite challenges.

Cassie vs. Jessica: A Tale of Two Leaders

In the short term, Cassie shines. Executives see the strategic roadmaps, the market insights, the polished presentations. Cassie gets promoted, rewarded, and recognized. The visible skills are impressive, and Cassie knows how to showcase them.

Jessica, however, works quietly. The invisible skills aren’t immediately apparent but play a crucial role in the long-term success of the product. Jessica foresees potential issues by understanding the product's place in the larger system. When problems arise, Jessica tackles them from the ground up, using first principles thinking. The team feels supported and motivated, thanks to Jessica's emotional intelligence and resilience.

Here's the paradox: Leaders with strong visible skills often climb the ladder quickly because these skills are easy to see and measure. They look good on paper and shine in presentations. But invisible skills? They’re harder to spot and even harder to quantify. Yet, they are the bedrock of true leadership.

A leader with excellent stakeholder management and presentation skills may win the day in a boardroom. But a leader with strong system thinking and first principles thinking will win in the long run. They’ll build products that not only succeed but also endure.

When we focus only on visible skills, we miss out on the deeper competencies that drive sustainable success. We end up with leaders who can impress in a meeting but struggle to steer the ship through turbulent waters. We miss the quiet leaders who think deeply, act wisely, and build lasting solutions.

Consider Cassie again. While impressive in the boardroom, Cassie often misses the underlying issues that Jessica sees clearly. Cassie’s projects face unforeseen challenges because the broader system implications weren't considered. The team, though impressed by Cassie’s presentations, feels the strain of unaddressed conflicts and high stress.

Jessica, meanwhile, anticipates and mitigates these issues. The product not only meets current demands but is also well-positioned for future challenges. The team thrives, feeling supported and understood. Yet, without visibility into these critical contributions, Jessica’s work often goes unrecognized by executive management.

The future of product leadership lies in balancing visible and invisible skills. Cultivate both. Recognize both. Promote continuous learning and self-awareness. Develop metrics that recognize the value of system thinking, first principles thinking, and emotional intelligence.

It’s about creating environments where invisible skills are nurtured and celebrated. Where leaders are encouraged to develop resilience, adaptability, and empathy alongside their strategic and technical abilities. This holistic approach to leadership development ensures that product leaders are not only effective in the present but also equipped to navigate future challenges.

Success in product leadership requires a blend of both visible and invisible skills. By acknowledging and nurturing these unseen competencies, organizations can unlock the full potential of their leaders. It's time to shine a light on the invisible skills that truly make a difference. The real drivers of innovation and sustained growth are often hidden in plain sight. Let’s bring them to the forefront.

True product leadership is about more than just what meets the eye. It’s about the depth of thinking, the ability to connect the dots, the resilience to keep going when things get tough, and the empathy to understand the needs of both your team and your customers.

These are the skills that, while invisible, are invaluable.

B2B SaaS is experiencing a seismic shift in UX strategies.

User Experience (UX) is no longer a secondary concern—it’s now the cornerstone of competitive advantage in B2B SaaS. In this evolving landscape, the focus has dramatically shifted from functionality to seamless, intuitive experiences. Unpacking the transformation, the strategies driving it, and the role of Product Managers in facilitating this change.

The Old Paradigm: Function Over Form

For years, B2B SaaS UX was all about functionality. Complexity was tolerated, even expected. Users were professionals; they could handle clunky interfaces if the software delivered value.

• Complex navigation: Users would navigate labyrinthine menus.

• Overwhelming feature sets: Tools packed with features but lacking intuitive design.

• Poor onboarding: Minimal guidance for new users.

The New Paradigm: Seamless and Intuitive

The shift is clear: today’s B2B users demand seamless, intuitive experiences. They expect their business tools to mirror the usability of consumer apps. This shift is driven by several factors:

1. Consumerization of Technology: Users now expect enterprise software to have the same UX standards as consumer apps.

2. Workforce Demographics: Millennials and Gen Z, who prioritize UX, are becoming a significant part of the workforce.

3. Competitive Landscape: UX is a key differentiator in an increasingly crowded market.

Key Strategies Driving the Shift

One of the fundamental strategies in this new paradigm is user-centric design. The user is at the heart of design decisions, ensuring that the software meets the real needs and preferences of its users. This involves creating user personas and journeys, which provide a deep understanding of user needs, pain points, and workflows. It also means adopting iterative design processes, where continuous feedback loops with real users help refine and improve the product.

Another crucial strategy is simplified onboarding. First impressions matter, and onboarding processes are being streamlined to ensure users can hit the ground running. Key components of this strategy include:

• Interactive tutorials: Guided experiences that help users learn by doing.

• Contextual help: In-app tips and support that provide assistance exactly when and where it’s needed.

Consistency and cohesiveness in the user interface (UI) are also vital. A consistent UI reduces cognitive load and increases user satisfaction. This is achieved through:

• Design systems: Standardized UI components and guidelines ensure a cohesive look and feel.

• Responsive design: Interfaces that work seamlessly across devices, from desktops to smartphones.

Data-driven UX is becoming the norm. By leveraging analytics and heatmaps, companies can understand user behavior and identify friction points. A/B testing allows them to experiment with different design elements to see what resonates best with users. This approach ensures that UX improvements are based on solid data and real user feedback.

Strategies for Existing Clunky Products

Transforming legacy products with poor UX can be challenging, but it's essential for staying competitive. One effective approach is incremental improvements. Small, iterative changes can significantly enhance the user experience without overwhelming users. This means:

• Prioritizing critical touchpoints: Focus on the areas where users experience the most friction.

• Gradual rollout: Implement changes in phases to allow users to adapt.

Refactoring complex workflows is another important strategy. Simplifying and streamlining existing processes can make them more intuitive. This involves:

• Mapping out user journeys: Identify and eliminate unnecessary steps.

• Introducing automation: Use automation to reduce manual effort and streamline operations.

Enhancing visual hierarchy is also crucial for improving the clarity and readability of the interface. Effective use of whitespace can create a cleaner, less cluttered interface, while highlighting key actions can guide users intuitively.

Strategies for Fresh Experiences

For new products, starting with a solid UX foundation is crucial. This begins with putting the user at the center of the design process. Essential steps include:

• Conducting comprehensive user research: Understand the target audience deeply.

• Creating detailed personas: Use these personas to guide design decisions.

Embracing minimalism is another key strategy. Less is often more in UX design. Focus on core functionality to ensure the primary use cases are exceptionally well-designed. Avoid feature bloat by introducing new features cautiously and ensuring they add real value.

Incorporating feedback loops for continuous improvement is essential for maintaining a high-quality UX. This can be achieved through:

• Beta testing: Engage users early in the development process.

• Regular surveys: Collect feedback consistently to inform future updates.

Thinking Framework for UX in AI

Designing UX for AI products introduces new complexities. Here’s how to approach it:

Start with the User

The technology you use should be guided by the user experience you want to achieve. Instead of diving headfirst into algorithms, think about how people do the task today. Figure out what’s valuable and how you can enhance the experience. Sometimes, a solution without AI might be easier to build or understand. When marketing, focus on user benefits rather than the AI technology.

Set the Right Expectations

Since everything from self-driving cars to smoothie makers calls itself AI, expectations are all over the place. People will expect your AI to be both smarter and dumber than it is. Explain in plain language what your AI can do and where its limitations lie. Generally, under-promising and over-delivering is a good way to build trust. Over time, users will learn how to best integrate the AI into their workflow.

Explain the Results

AI is only useful if users understand its decisions. Ideally, users should be able to trace any result back to supporting data points. If that’s not possible, explain the basic operation of the algorithm. Lay out which data sources you use and which qualities the AI focuses on. If you aggregate data from multiple sources, break them down to let the user reproduce the result. This information should be available as part of the user flow through a consistent interface.

Communicate Your Confidence

Users rely on your AI to make decisions. They need to understand the quality of results to trust them. If the confidence of your algorithm varies, indicate the confidence for each result. You could show a percentage or use an abstract visualization (e.g., star ratings, colored indicators). For results with multiple parts, break down the confidence for each. Additionally, consider showing multiple results ordered by confidence and giving the user the final say.

Degrade Gracefully

Designing for AI means designing for many different outcomes. When the input is clear and the answer certain, you don’t want the user to hesitate. Less confident results need to be presented differently. You could start by toning down the boldness of your visual design or altering the layout and copy that frames the result. Above all, don’t be afraid to say when you don’t have an answer. It’s okay for an AI to fail, as long as you design for it.

Know What Not to Automate

Not everything should be automated. Some tasks require abilities unique to humans, such as understanding emotions or motivations, providing dignity or enjoyment, or making ethical or moral decisions. Other tasks may have far-reaching consequences for an affected party. Identify which parts of a task are suitable for AI and which should remain human-driven.

Keep the User in Control

Instead of an AI that replaces humans, think of ways to amplify and augment human abilities. The user should always be in control, able to intervene, provide feedback, reverse bad actions, and reward good ones. AI is more empowering when it works with the user, not for the user.

Build Trust Over Time

Be careful when introducing your AI to new users. Ensure it doesn’t require much existing personal data initially. Make suggestions instead of decisions. As your AI gets to know the user, you can automate more and ask for permission less. This gradual approach helps users understand how the AI works and allows the AI to learn alongside them.

Help Your Users Grow

Over time, the AI will need to adapt to changing user behavior. This applies to individual users and the user base as a whole. Ensure your AI evolves with changing societal values and needs, aiming ahead of the curve without imposing your values on users.

Balance Predictability and Serendipity

Any personalized AI adopts the user’s bias, which is great for tasks requiring predictability. However, it can limit curiosity and constrain users to their comfort zones. Design your algorithms to balance predictability with the possibility of discovering something new, and offer ways for users to escape the filter bubble.

Product Managers (PMs) are going to play a pivotal role in driving the UX revolution in B2B SaaS. Prioritizing UX in the product roadmap is essential. PMs need to strike a balance between new features and UX enhancements. Aligning UX improvements with the long-term vision of the product ensures that the user experience is consistently prioritized and improved over time.

One product I really admire that’s driving the shift in UX thinking from front, is Command Bar.They are a SaaS in user assistance category. They believe User assistance shouldn’t be passive. It needs to be proactive, adaptive, and intuitive—like having a personal software butler who knows exactly what you need and when you need it.

Listening to User Intent:Most users know what they want to accomplish, but the software often fails to capture this intent. Imagine a user session where someone is clicking around in frustration, unable to complete a simple task. Enter CommandBar's 'Copilot,' a natural language interface where users can chat with an AI-powered assistant. This assistant does more than just answer questions; it completes actions and triggers walkthroughs based on user intent.

Consider this scenario: a user wants to set up a Salesforce integration. Instead of digging through menus, they simply ask Copilot, "How do I set up Salesforce integration?" and get a direct response, "Let's set up the integration—what’s your Salesforce account ID?" This isn’t just helpful—it’s transformative. It shifts the software experience from reactive to proactive, addressing user needs as they arise.

Intervening at the Right Moments:

We’ve all been there—clicking around in confusion or hitting the same button repeatedly out of frustration. Traditional software either ignores these signs or offers generic help popups that often miss the mark. CommandBar’s 'Nudge' system is different. It intervenes at just the right moments, offering contextual help based on user behavior.

For instance, if a user seems confused or is rage-clicking, a nudge might pop up saying, "Need help with this feature?" or "It looks like you’re trying to do X, here’s how to do it." This kind of timely intervention can be a game-changer, preventing user frustration and enhancing overall experience.

One size doesn’t fit all, especially when it comes to user assistance. CommandBar understands this, offering personalized interactions that adapt to each user's needs. New users might receive detailed guidance, while experienced users get the freedom to explore.

Popups have a bad reputation for being annoying and intrusive. CommandBar aims to change that. Nudges are not indiscriminate popups; they are carefully targeted messages that appear based on user behavior and past interactions. For example, showing a nudge about a new feature to users who previously asked about it builds trust and shows that the software is genuinely trying to help.

CommandBar is more than just a tool; it’s a vision for the future of software interfaces. As AI continues to evolve, the potential for more intuitive and adaptive user assistance grows. CommandBar envisions a world where interfaces dynamically adjust to user needs, providing help exactly when and where it’s needed.

We might be witnessing the end of traditional web design. Interfaces need to evolve to keep up with changing user expectations. Products like Command Bar, Superhuman and Arc inspire us to rethink what software can be.

The UX landscape for B2B SaaS is evolving rapidly. Companies that embrace this shift and prioritize user-centric design will not only meet the demands of today’s users but will also set themselves up for long-term success. Great UX is no longer a nice-to-have; it's a must-have. The future of B2B SaaS lies in experiences that are as intuitive and engaging as the best consumer apps.

Responsibilities that once defined the role are now scattered across various teams. This dilution threatens the core of what it means to be a product manager. If we want to remain relevant and effective, we must reclaim our essential duties and blend commercial acumen with technical expertise.

The Fragmentation of Responsibilities

Today’s product management landscape is fragmented. Key functions traditionally managed by PMs are now owned by different teams:

• GTM Strategy: Managed by marketing and other go-to-market teams

• Pricing and Packaging: Controlled by finance or strategy teams

• Growth: Driven by growth teams.

• Product Vision: Defined by Executives.

• Product Strategy: Crafted by strategy teams with management consulting backgrounds.

• Business Modeling: Owned by research and analytics.

• Customer Voice: Managed by customer success teams.

• Product Analytics and Operations: Managed by specialized operations teams.

Everything that was once integral to product management has been divided, leaving PMs narrowly focused on delivery and execution, stripping the role of its strategic and commercial essence.

I strongly see the need for General Managers as Product Leaders who drive business outcomes with the product. Today's PM role is ineffective.

This was however the classic PM role from before. In tech, we’ve strayed from this model. PMs must return to being commercially savvy. But if we start calling them 'GMs,' they need true GM authority, or we'll repeat the 'CEO of the product without being a CEO' problem.

The heat is on for entrepreneurial leaders, those who understand the full spectrum of bringing a product to life and ensuring its financial success. The pressure is mounting to not just manage but to own the business outcomes of our products.

The Historical Context

Product management hasn't always been the fragmented role we see today. Its origins lie in a far more integrated approach where the product manager was the lynchpin of both commercial and technical aspects of a product.

The Evolution of Product Management:

• Origins: Initially, product management was a commercial function focused on business success and market outcomes. PMs were expected to have a deep understanding of market dynamics, customer needs, and business strategies. They were the bridge between the market and the engineering teams, ensuring that the product not only met technical specifications but also aligned with market demands and business objectives.

• Shift to Technical Delivery: Over time, as products became more complex and technology-driven, the role of the PM began to shift towards technical delivery. PMs started focusing more on feature specifications, sprint planning, and delivery timelines, while commercial aspects were handed over to specialized teams. This shift created a void in commercial responsibilities, which were increasingly taken over by marketing, finance, and strategy teams.As the focus on technical delivery intensified, specialized roles emerged to handle various aspects of product management. These roles included data analysts, customer success managers, growth hackers, and more. Each of these roles took over a slice of the PM’s original responsibilities, further fragmenting the role and diluting its impact.

The Path Forward

To address this existential crisis, PMs need to reclaim their core responsibilities and integrate commercial acumen with their technical expertise.

A) PMs Need to Become Commercially Savvy Again:

PMs must deepen their understanding of market dynamics, financial metrics, and business strategies. This involves:

• Market Analysis: PMs need to stay abreast of market trends, competitive landscape, and customer needs. They should conduct thorough market research and competitive analysis to identify opportunities and threats. Understanding market dynamics is crucial for making informed decisions about product strategy and positioning. This piece is mostly still happening in product teams.

• Financial Acumen: PMs must master pricing strategies, revenue models, and cost structures. They should understand how their products generate revenue and what the key cost drivers are. This knowledge is essential for making strategic decisions about product features, pricing, and go-to-market strategies. PMs should also be able to build and analyze financial models to evaluate the business impact of different product strategies.

• Strategic Thinking: Developing product strategies that align with business goals and drive growth is essential. PMs should be able to translate business objectives into product strategies and roadmaps. They should also be able to communicate these strategies effectively to stakeholders and align cross-functional teams around a common vision. Strategic thinking involves not only setting the direction for the product but also being able to pivot and adapt to changing market conditions.

B) Organizations Must Leverage PMs Effectively:

Companies need to recognize and fully utilize the capabilities of their PMs by:

• Supporting Development: Investing in training to enhance PMs' commercial and strategic skills. Companies should provide PMs with opportunities for continuous learning and development. This can include training programs, workshops, and mentorship opportunities. By investing in the development of their PMs, companies can ensure that they have the skills and knowledge needed to drive product success.

• Empowering PMs: Providing PMs with the authority and resources to make strategic decisions. This includes giving PMs the autonomy to prioritize features, allocate resources, and make trade-offs based on business objectives. Empowered PMs are more likely to take ownership of the product and drive it to success.

Product management is at a crossroads. To navigate this existential crisis and role encroachment in tech, PMs must reclaim their core responsibilities, embrace strategic thinking, and act as true owners of their products.

Most founders obsess over it. Investors demand it. But what exactly is it?

PMF isn't just about creating a great product. It’s about building the right product for the right market. It requires understanding market needs deeply and iterating constantly based on feedback. A product that resonates so well with its audience that it practically sells itself.

Understanding Your Market

First, understand your market. Not just on the surface. Dive deep.

Conducting thorough market research is the foundation of achieving PMF. This involves employing various techniques like surveys, interviews, and data analysis to gather comprehensive insights about your target audience. Surveys can provide quantitative data, while interviews offer qualitative insights. Data analysis, including trend analysis and segmentation, helps understand different customer segments.

Understanding your audience's pain points and unmet needs is crucial. This means going beyond superficial observations and uncovering the deeper issues your product needs to solve. Surface pain points are the obvious issues users face, while deep pain points are underlying problems that may not be immediately apparent. Empathy mapping helps visualize users' feelings and thoughts, while customer journey mapping tracks their interaction with the product over time.

Misinterpreting feedback can lead to addressing symptoms rather than root causes. Overlooking subtle signals might mean missing critical insights. Engage regularly with users through forums, user groups, and direct interactions. Combine qualitative insights (e.g., interviews) with quantitative data (e.g., usage analytics) to get a holistic view.

Remember, you’re not just selling a product. You’re solving a problem. The bigger the problem, the more desperate the need for your solution.

Iterative Development

Next, embrace iterative development. The days of long development cycles and big reveals are over.

Today, it’s all about build, measure, learn. Launch a minimum viable product (MVP). Collect feedback. Analyze trends. Repeat the process. Again and again.

This approach isn't without its challenges. Limited resources might restrict the scope of experiments. Balancing speed and quality can be tricky; too fast, and you might overlook important details; too slow, and you miss market opportunities. Prioritize features that offer the most value.

Collecting feedback is a crucial part of this process. User testing provides direct observations, beta launches gather early adopter insights, and surveys collect structured responses. However, high volumes of feedback can be overwhelming, and contradictory opinions can create confusion about which direction to take. Categorize feedback to identify common themes. Focus on recurring issues that impact a broad user base. Use tools like heatmaps and session recordings to understand user behavior.

Sometimes, despite your best efforts, you may need to pivot. Recognizing when and how to pivot is essential. Signs that a pivot is necessary include stagnation in user growth, high churn rates, and consistent feedback about unmet needs. Team resistance often comes from emotional investment in the current vision. The sunk cost fallacy makes it hard to abandon significant investments. Make data-driven decisions to justify pivots. Communicate transparently with the team about the reasons and expected outcomes. Celebrate small wins post-pivot to build morale.

Quantifiable Indicators

How do you know when you’ve hit PMF? Look for quantifiable indicators.

High user engagement. Users actively using your product, integrating it into their daily routines. Metrics like Daily Active Users (DAUs) and Monthly Active Users (MAUs) provide a snapshot of user activity. Session duration indicates how much time users spend on your platform. Vanity metrics (like total downloads) can be misleading if not correlated with actual engagement. Misinterpreting data can lead to incorrect conclusions. Focus on metrics that reflect genuine engagement and retention. Use cohort analysis to track user behavior over time. Implement A/B testing to understand the impact of changes.

Low churn rates. Customers sticking around, finding ongoing value. Retaining customers is often more cost-effective than acquiring new ones. High retention rates signal a strong product-market fit. Identifying the root causes of churn can be complex. It might be due to product issues, poor customer support, or external factors. Conduct cohort analysis to understand when and why users churn. Use exit surveys to gather insights directly from departing customers. Implement retention strategies like personalized onboarding and regular check-ins.

Revenue growth is a good sign, but it’s not the only metric. Customer satisfaction is crucial. Happy customers are your best advocates. Balancing growth and satisfaction is key. Rapid revenue growth is positive, but not if it comes at the expense of customer satisfaction. Unsatisfied customers will eventually churn. Overemphasis on revenue can lead to neglecting user experience. This short-term focus can damage long-term sustainability. Regularly measure NPS to gauge customer satisfaction. Prioritize user feedback in product development. Create a customer success team focused on maintaining high satisfaction levels.

Addressing Complexities in PMF

Achieving PMF is not without its challenges. It requires navigating a landscape of market dynamics, internal challenges, and external factors.

Market Dynamics

The market is always evolving. New competitors emerge, user preferences shift, and technological advancements create new opportunities and challenges. Rapid changes can make it hard to stay relevant. New competitors can disrupt your market share. Conduct continuous market research to stay updated. Develop agile response strategies to quickly adapt to changes. Foster a culture of innovation to stay ahead of competitors.

Internal Challenges

Achieving PMF requires a unified effort from the entire team. Everyone needs to be on the same page regarding goals and priorities. Different departments might have conflicting priorities. Communication gaps can lead to misunderstandings and misaligned efforts. Hold regular meetings to ensure alignment. Set clear goals and KPIs that everyone understands and works towards. Foster a collaborative culture where feedback and ideas are freely shared.

External Factors

External factors like economic downturns can impact customer behavior and purchasing power. Unpredictable shifts can create budget constraints and reduce customer spending. Diversify your customer base to mitigate risks. Develop adaptive pricing models that can cater to different economic conditions. Stay agile and be prepared to adjust your strategies as needed.

Learning from real-world examples can provide valuable insights into the journey towards PMF.

Slack

Slack started as a gaming company but pivoted to a messaging platform after recognizing a bigger market opportunity. Their journey highlights the importance of user feedback and the flexibility to change direction when necessary.

Initially, Slack’s product, a game called Glitch, failed to gain traction. However, the team noticed that their internal communication tool was highly effective. They decided to pivot and develop the communication tool into a product, which became Slack. This pivot was driven by user feedback and an understanding of the market’s needs.

Airbnb

Airbnb faced trust issues from users who were hesitant to stay in strangers' homes. Their journey underscores the need for a deep understanding of user needs and innovative solutions to build trust.

Initially, Airbnb struggled to gain users and faced skepticism about the safety and reliability of their service. They introduced features like user reviews, verification processes, and a host guarantee to build trust. This deep understanding of user concerns and innovative solutions to address them were crucial in achieving PMF.

Superhuman

Superhuman, the email client known for its speed and efficiency, offers a fascinating example of using structured feedback to achieve PMF. Their approach involved the “very disappointed” survey, a method developed by Sean Ellis.

Superhuman asked users, "How would you feel if you could no longer use Superhuman?" with options ranging from "Very disappointed" to "Not disappointed." This survey aimed to identify how essential the product was to its users. The goal was to have at least 40% of users respond that they would be "very disappointed" if they could no longer use the product.

When Superhuman initially conducted this survey, they found that less than 40% of users would be very disappointed without the product. Instead of being disheartened, they used this feedback to dig deeper. They asked follow-up questions to understand what features users valued most and what pain points remained.

Armed with this data, Superhuman iterated on their product. They focused on refining the features that users loved and addressing the concerns that prevented others from being very disappointed without it. Over time, through continuous iterations and improvements, they finally achieved the desired 40% threshold.

This method allowed Superhuman to pinpoint exactly what their users needed, helping them tailor their product to fit the market perfectly. It’s a powerful example of how structured feedback and targeted improvements can lead to PMF.

Achieving PMF is not a one-time event. It’s a continuous journey of adaptation and alignment. Stay vigilant, listen to your market, and adapt as needs and trends evolve. The journey to PMF is arduous but rewarding. It’s the foundation of sustainable growth and long-term success.