The old Product Manager was a clerk.

They gathered requirements, wove roadmaps, and acted as a shield for the engineering team—keeping the “real world” at bay so the work could get done.

To lead—to actually make a dent in the new AI world—the PM must become an entrepreneur. While everyone else is chasing the hype, the entrepreneur creates meaning.

In the old world, the roadmap was the product.

If you followed the roadmap and shipped on time, you won. But in the world of AI, the map is noise. Shipping a feature or a model isn’t the same as creating value.

The PM in the world of AI can’t afford to be a bureaucrat. They have to be an entrepreneur. They aren’t managing a department; they are running a venture inside a fortress.

Here is how the entrepreneur survives the shift:

The Cold Start

An entrepreneur doesn’t wait for a market to appear; they go out and prove one exists.

Before we write a single line of production code, we have to find the truth. Does the data actually have something to say? Can the model solve a pain point that matters? If we can’t prove the value with a prototype and a prayer, a million dollars of budget won’t save us.

The Uncertainty of “Done”

In the old world, a login button worked. We built it, it stayed built.

AI isn’t like that. It’s non-deterministic. It’s 90% accurate on Tuesday and 85% on Thursday because the world changed while we were sleeping. This isn’t a bug; it’s market volatility. As entrepreneurial PM, we don’t whine about “data drift”—we manage the uncertainty. We build systems that are resilient to a world that won’t sit still.

Falling in Love with the Problem

The amateur falls in love with their solution. They cling to their LLM or their specific architecture like a life raft.

The entrepreneur knows better.

If the “best” way to solve the problem changes every ninety days, we need new models, new math, new possibilities. If we’re wedded to our approach, we’re already obsolete.

Don’t fall in love with the tool. Fall in love with the customer’s struggle. Be willing to pivot, to scrap, and to reinvent—as many times as it takes to be remarkable.

The roadmap is dead. The mission is all that’s left.

Here is how we build for it:

Distinguish the Signal from the Noise

Most people are currently drowning in the AI noise.

The “AI-powered” sticker slapped onto a legacy feature. The frantic race for the longest prompt or the newest, shiniest model. It’s loud, it’s expensive, and mostly hollow.

The Signal, however, is the core capability that actually changes the customer’s day.

As an Entrepreneur PM, our job isn’t to follow the parade. It’s to find the signal. We have to be brave enough to ask: What is this for? If the technology doesn’t solve a friction point that was previously unsolvable, we aren’t innovating. We’re just decorating.

An entrepreneurial PM looks at a loud market and ignores the glittering sirens. Instead, they look for the one quiet, powerful signal that has the potential to create a “wow” moment.

Are you adding to the noise, or are you building the signal?

The PM as a High-Frequency Investor

In the old world, we made “smart bets” once a quarter. In the AI world, the risk curve is upside down. Building is cheap; knowing what to build is expensive.

We are no longer just a manager of resources; we are an investor of capital (engineering hours, design cycles, focus, and our customer’s trust).

The entrepreneurial AI PM shifts from quarterly roadmap planning to high-frequency experimentation. We aren’t looking for one big win; we are looking for the signal in a sea of data. We invest small amounts of “capital” into multiple hypotheses, doubling down only when the customer behavior confirms the value.

Innovate in the Constraints

The legacy logic says we need a massive budget and a dedicated ML team to innovate. The Entrepreneur PM says: “What can I do with what I have right now?”

Innovation isn’t about having the best model; it’s about having the best insight.

Key Parallels Between the PM & the Entrepreneur

The roles converge on the concept of Value Discovery rather than just Feature Delivery.

• Vision and Evangelism: Both must sell a “future state” that isn’t yet visible. Whether it’s convincing a VC to invest or an Engineering Director to commit resources, both roles require high-level persuasion based on a hypothetical outcome.

• Resourcefulness (The “Scrappy” Factor): AI development is expensive and resource-constrained. An AI PM, like an entrepreneur, must find creative ways to get “labeled data,” compute time, or “design partners” (early users) without a massive initial budget.

• Obsession with Product-Market Fit: For an AI PM, this is often “Model-Problem Fit.” They spend a disproportionate amount of time validating if the AI’s output actually changes user behavior or creates a business advantage.

How They Are Alike vs. How They Differ

While the mindsets are similar, the operating environments create distinct differences.

The Core Differentiator: The “Safety Net”

The fundamental difference is the infrastructure of support. An AI PM has the luxury (and the burden) of an existing brand, a sales team, and legal frameworks. The Entrepreneur must build those frameworks while simultaneously building the product.

However, the most successful AI PMs ignore the safety net. They treat their product’s “compute budget” as if it were their own money and their “user’s trust” as if it were their own reputation. That is the essence of the Entrepreneurial PM.

Share

Leave a comment

Most people think a Product Manager is a manager. They aren’t.

The best ones? They’re investors.

They wake up every morning with a finite bag of chips—engineering hours, design cycles, and focus— and they have to decide where to place them on the table. They don’t have a map; they only have a compass and a hunch.

If you spend your “capital” on a feature that no one misses, you’ve lost more than time. You’ve lost the chance to make a difference.

Strategic product work isn’t about checking boxes or clearing tickets. It’s about the brave act of allocating scarce resources in the face of the unknown. It’s about seeking a return that matters.

The question isn’t “What can we build?” The question is “Where will our investment create the most change?”

Both roles are essentially “allocators”

• The Investor allocates financial capital (cash) across a portfolio of assets (stocks, startups, real estate).

• The PM allocates “R&D capital” (engineering hours, design cycles, and budget) across a portfolio of features or initiatives.

In both cases, saying “yes” to one opportunity inherently means saying “no” to another (Opportunity Cost). A PM must decide if a “big bet” on a new market is worth the cost of not fixing technical debt.

Neither an Investor nor a PM has a crystal ball

• Due Diligence: An investor grills founders and checks financials; a PM conducts user discovery and market research to validate a hypothesis before committing resources.

• Hedging: An investor diversifies to protect against a total loss. A PM might run a “painted door” test or a small Beta to limit the “downside” of building a feature that nobody wants.

Portfolio thinking

A sophisticated PM doesn’t just look at one feature; they manage a Product Portfolio with varying time horizons:

• Core Assets (The “Cash Cows”): Reliable features that keep the lights on and users retained.

• Growth Bets: High-potential initiatives that could significantly move the needle.

• Moonshots (The “VC” Bets): High-risk, high-reward experiments that might fail but could define the company’s future.

The Goal: Return on Investment.

The “Return” is simply measured differently:

• The Investor looks for Alpha (beating the market) and Internal Rate of Return (IRR).

• The PM looks for Product Alpha—delivering value that exceeds the cost of development and outperforms competitors. This is measured through KPIs like LTV (Lifetime Value), CAC (Customer Acquisition Cost) efficiency, and retention.

How the Mental Model Shifts with AI

From Capital Allocation to “Agent Orchestration”

In the traditional model, your capital was limited by the number of engineering hours in a sprint. With AI-assisted coding and automated testing, the “cost per feature” is dropping.

• The Shift: You are no longer just an investor in human talent; you are an orchestrator of a hybrid workforce. You must decide which tasks are “sub-contracted” to AI agents (e.g., initial PRD drafts, feedback synthesis, QA) and which require the high-cost “artisan” touch of your senior developers.

From Static Due Diligence to “Continuous Monitoring”

Traditional investors perform due diligence at a point in time (e.g., a quarterly review). AI allows for always-on awareness.

• The Shift: Instead of waiting for a monthly report to see if your “investment” in a feature is paying off, AI tools now provide real-time sentiment analysis and anomaly detection. You can “cut your losses” or “double down” on a feature in days rather than months.

The “Alpha” of Proprietary Data

In investing, “Alpha” comes from information others don’t have. In an AI world, everyone has access to the same LLMs.

• The Shift: Your edge no longer comes from being able to summarize data (AI does that); it comes from the proprietary data pipelines you build. As a PM, your “investment moat” is now the unique, high-quality user data your product captures that competitors can’t replicate.

Essential Skills for the Future (2026 & Beyond)

As the “mechanical” parts of product management (writing tickets, documentation, basic data pulling) are commoditized, these three skill pillars will define the top 1%:

AI & Data Literacy (The “Quant” Skill)

• Prompt Engineering as Literacy: Moving beyond “chatting” to building agentic workflows. You need to know how to structure inputs so an AI can execute a multi-step process (like market research + competitive analysis + SWOT).

• Understanding “Model Drift” & Bias: Just as an investor watches for market volatility, you must understand when your AI features are becoming less accurate or reflecting biases that could alienate users.

Systems & Lifecycle Thinking

• Managing MLOps: AI products mature differently. They require continuous training loops, not just “ship and forget.” You need to understand the lifecycle of a model—from data cleaning to inference costs.

• ROI Modeling with AI: Using AI to simulate outcomes. In 2026, top PMs use “synthetic users” or simulation models to test a feature’s potential ROI before a single line of production code is written.

Human-Centered Strategy & Ethics

• The “Empathy Moat”: AI can predict what a user will do, but it often struggles with why. Your ability to conduct deep, empathetic user interviews remains the ultimate “un-AI-able” skill. This is where you go back to being The Behavior Architect and you focus on The Architecture of Trust.

• Responsible AI Governance: Understanding the legal and ethical landscape (e.g., the EU AI Act). Being a “fiduciary” for your users means ensuring their data is safe and the AI’s decisions are explainable.

Send us a message to learn more about The Quantum Collective upcoming round-table. What topics or areas would you like to learn more about? Tell me more about your journey as AI Product Manager?

Do you want to test your AI Product Readiness? Try our Free Assessment Tool.

Here is a paradox of the AI era: Most people are being trained by their models. They take what the LLM gives them, tweak it 5%, and ship it. They are becoming the apprentices to their own tools. They are being trained to think in the “averages” of the available data.

We talked about the evolution from Product Manager to Behavioral Architect as the necessary move from the "What" and the "How", to the “Why”. But as the AI engine gets louder and the noise begins to drown out reason, even the Architect finds themselves at a crossroads. What is the way?

We used to optimize for clicks. Then we optimized for behavior. But how is this different from optimizing for ‘conversions’?

Solving the “Rationality Gap”

AI is perfectly rational, but our users are not. AI can tell us that a user should click a button based on a million data points, but it cannot feel the Defensive Anxiety, or the Status-Seeking, or the Need to Belong, that actually drive the human hand. This is why the PM-as-Psychologist becomes necessary.

• The PM-as-Psychologist understands that people don’t buy products, or features; they buy a “better version of themselves.” A result. A benefit. A transformation. A feeling.

• The PM-as-Psychologist understands that 15% of the buying decisions are determined by the past, while 85% are determined by the anticipated future: “What is goint to happen as a result of me buying?”

• The PM-as-Psychologist knows and applies models like Loss Aversion or the Endowment Effect to understand why a user might frown at (or fear) the new UI, even if it’s technically superior.

In the era of “Quantum” Excellence, our role as PMs shifts from managing the backlog to managing the psyche.

As AI takes over the technical “doing”—writing user stories, analyzing data, and optimizing flows—our value as PMs shifts to the messy, irrational, and deeply human elements of building product through Trust.

The Architecture of Trust

In a world of automated noise, Trust is the only non-commodity. You can’t prompt your way into a relationship.

Thinking like a psychologist allows you to integrate the polarities of technology and humanity. It allows you to stop the “Panic-Stagnation” and move towards excellence—not by working faster, but by removing the psychological friction that slows everyone else down.

And then you optimize for Trust. Trust is one of the few currencies that AI cannot mint. AI can simulate empathy, but it cannot be empathetic. A PM-as-Psychologist applies psychology models—not just to lead a user to a button, but to build a bridge of radical reliability.

Use models like the Trust Triangle (Authenticity, Logic, Empathy) to audit every feature.

• The Check: Does this AI-generated response sound like a machine trying to be human (uncanny valley), or a human using a machine to be more helpful?

• The Quantum Shift: When you stop optimizing for engagement and start optimizing for psychological safety, your product’s brand shifts from predatory to partner.

The PM-as-Psychologist understands that the customer only buys improvement. And the reason they buy is that they feel the quality and quantity of that improvement is greater than the cost.

Leave a comment

The Other Side: Managing the “Inner Engine” of the Team

High-performing teams don’t fail because of bad code. They fail because of bad psychology.

The PM-as-Psychologist understands that the team is its own product. They audit the Inner Graffiti of the group. (for more on that check The Excellence Habit). They recognize when anxiety is leaking into the roadmap. By applying psychological safety models (like Amy Edmondson’s), they create an environment where a “failed” experiment is seen as a data point, not a death sentence.

Efficiency isn’t about more Jira tickets per week. It’s about the velocity of trust. Inside the team and in the marketplace, where the product lives.

Product development can be a high-pressure environment, which often leads to The Great PM Anxiety and the Always-on culture.

Here is the glitch: When a team is stressed, they fall into “Feature Factory” mode—shipping things just to feel productive, even if those features add no value.

• It is upon the PM-as-Psychologist to create psychological safety. They know that when team-members are afraid to fail, innovation fades. By managing the team’s collective temperament, the PM ensures the “Inner Engine” is humming, not just spinning its wheels.

Training the Mirror (Not the Other Way Around)

Let’s go back to the paradox of the AI age: Most people are being trained by their models.

Unless you start thinking like a psychologist, the AI models will eventually train you. So here is the playbook:

• Accept your role of PM-as-Psychologist, and treat AI as a mirror of your own intent.

• Train the model by injecting your unique identity—your values, your ethical filters, and your understanding of human nuance. Become the teacher, not the student, of the algorithm.

• Don’t ask the model what to do. Tell the model who you are trying to serve and why.

• Train the AI to be your Force Multiplier, not your replacement.

Summary: The Quantum Excellence Shift

Share

AI is a world-class tool. An engine. It can generate 1,000 user stories before you finish your morning espresso. It can optimize a checkout flow until the friction is theoretically zero. It can predict what a user might do based on what a million other people did yesterday.

But AI has a blind spot: it is perfectly rational. And humans are not.

AI sees data; the PM must see meaning. AI hears the noise; the PM must find the signal.

Beyond Product Management: The Rise of the Behavioral Architect

The industry is undergoing a silent transition. We used to be Product Managers—custodians of backlogs, roadmaps, and shipping dates. Then we became Decision Architects, framing choices to lead users toward a “conversion.” In my previous post - The Engine and the Noise, we make the case that transitioning from a PM to a Decision Architect solves for the paradox of needing slower thinking, in an age where AI accellerates everything else.

The next step in this evolution calls for more. We must become Behavioral Architects.

A Behavioral Architect doesn’t just build a bridge; they understand why a human might be afraid to cross it, even if the math says it’s safe. AI can optimize for “clicks,” but it can’t feel the “irrationality” of a human user’s desires, fear, status-seeking, or need for belonging.

The Tool: The “Inner Graffiti” Audit

In The Excellence Habit, we learned about the internal noise that clouds our judgment. In product development, this manifests as Inner Graffiti—the ego-driven development that muddles a clear product vision. To gain clarity and evolve towards a behavioral architect, we need to wipe this internal noise and clear the ‘inner graffiti.

Here is how to run the “Inner Graffiti” Audit. Ask your team:

• Are we building this because the data suggests a real “gap,” or because we want to demonstrate we are “in charge” of our product? (Ego-driven development).

• Is this feature a response to a competitor’s move, or a response to a user’s unspoken pain?

• Are we building this for the customer, or for our job-security?

• Did we explore reasonable alternatives? Did we ‘talk’ to the customer personas we created?

The Litmus Test: If no one ever knew we built this, would it still be the right thing to do?

If the answer is about us rather than the customer, you aren’t building a product; you’re just spraying more of the inner graffiti on the user’s experience.

Ethical Wielding of the Behavioral Toolkit

Behavioral Science provides us with powerful, almost dangerous, tools.

• Loss Aversion: We know users hate losing $10 more than they love gaining $10.

• Choice Overload: We know that 24 types of jam lead to fewer sales than six.

AI can weaponize principles like these to drive short-term metrics. It can use Loss Aversion to “trap” a user into a subscription. It can use Choice Overload to paralyze them into staying on a platform. Did we serve the customer when we implemented those?

This is where “Inner Excellence” meets the roadmap. A Behavioral Architect uses these tools with a clear focus. They wield Loss Aversion to help a user protect their health or their savings. They use Choice Overload reduction to give a user back their time, not take more of it away.

The Shift

The engine (AI) will only get louder. It will produce more noise, more stories, and more “optimized” features that deliver on metrics, but feel hollow.

Your job is no longer to compete with the engine. Your job is to be the architect who ensures the engine is driving toward something that matters. Something that is meaningful to your customers, and your business.

AI optimizes for the how. You must optimize for the why.

Don’t just manage a product, architect a behavior.

The industrial era was about the machine. The digital era was about the code. But the “Quantum” era? It’s about the vibration of the person behind the product.

Most Product Managers are busy building “to-do” lists. They are adding features, adding meetings, and adding stress. They think that by doing more, they will be more. They are wrong.

Excellence isn’t an additive process. It’s a subtractive one. It’s about clearing the “Inner Graffiti” so the engine can finally hum. If you want to move from “Senior PM” to “Epic Life,” you don’t need a new Jira board. You need a new inner engine.

Here are the three “quantum” habits that power it:

1. Cultivate a “Digital Temperament”

The lizard brain hates the unknown. When a new AI tool drops, the default reaction isn’t curiosity—it’s defensive anxiety. We ask, “Will this replace me?“ instead of “How can this expand me?“

Digital Temperament is the psychological habit of approaching the “new” with resilient curiosity. It’s a muscle. You don’t build it by reading manuals; you build it by breaking things.

• The Practice: Run one low-stakes “AI experiment” every Tuesday (or Wednesday … pick your day).

• The Quantum Shift: Stop the defensive “But.” Integrate the tool as an extension of your intent, not a threat to your identity.

How to do this? James Clear’s laws for building new habits state:

1. Make it Obvious: Set a weekly “AI Experiment” calendar invite for Tuesday mornings (or pick your day) and place a shortcut to a new LLM or tool directly on your desktop.

2. Make it Attractive: Join a community (like the Quantum Collective) or find a “curiosity partner” where the social reward for sharing a failed experiment is higher than the reward for being right.

3. Make it Easy: Commit to “One Prompt a Day”—a two-minute interaction where you ask an AI to explain a concept or draft a simple email, lowering the friction of exploration.

4. Make it Satisfying: Create a “Wins & Wonders” log where you track one task that became 10% faster or one “Aha!” moment found through a tool each week.

2. Metacognition Over Execution

We are addicted to the “Done” list. We execute because execution feels like safety. But in a world of automated intelligence, “doing” is a commodity. “Thinking about the doing” is the premium.

High-performers are shifting the question from “How do I do this?” to “How should I think about this?”

• The Audit: Look at your last three big decisions. Were you chasing the Status Quo? Did you let an AI output bypass your critical filter because you were in a rush?

• The Paradox: You have to slow down to move at quantum speed. Auditing your decision-making process is the only way to stop swinging between the extremes of “Panic” and “Stagnation.”

To perform “The Audit” and resolve “The Paradox,” you must shift from enabling a “Feature Factory” to becoming a “Decision Architect.”

Here are four principles that serve as the mental scaffolding for this transition. These principles will allow you to stop the “Panic-Stagnation” swing and move at Quantum Speed.

1. The Superposition Principle: “Am I deciding too early?”

2. The Noise-to-Signal Filter: “Is this AI output a foundation or a finished product?”

3. The Friction Audit: “Where am I trying to ‘win’ instead of ‘solve’?”

4. The Entanglement Test: “How does this decision impact the whole system, not just the feature?”

A detailed examination of these principles is coming in a follow-up post. Sign-up for free so you don’t miss it.

3. The “Fresh Start” Ritual

Ego and doubt are like graffiti on a train. If you don’t wash it off, eventually, you can’t see out the windows.

Behavioral science tells us that “threshold moments”—Mondays, the first of the month, the kickoff of a new project—are windows of high plasticity. These are the moments when the “old you” is most willing to step aside.

• The Ritual: Use every Monday morning to reset your Inner Engine.

• The Integration: Don’t fight your doubt; acknowledge it, then overflow it with the light of your current mission. Clear the ego. Start fresh.

The Bottom Line

You aren’t here to fight your shadow or outrun the machines. You are here to integrate the two.

When you stop the wrong habits—the winning too much, the constant execution, the defensive posturing—something magical happens. Balance. Not the boring balance of neutrality, but the Power in Harmony.

That is the Quantum Habit, the new product management. That is how you vibrate in wholeness.

Stop doing. Start being. Then ship.

Speed is no longer a competitive advantage.

If you can prompt it, so can your competitor. If an AI can roadmap it, the value of that roadmap just hit zero.

We are entering the era of the Human Moat. The “next-gen” Product Manager isn’t the one who uses AI the most; it’s the one who knows exactly where the machine stops and where human judgment must begin.

To win in 2026, you don’t need a faster processor. You need a better lens. Here are three frameworks to build your moat.

1. The Behavioral Architect: Designing for the Heart

As PMs we all have the tendency to treat features like a grocery list. We check boxes and wonder why the user isn’t buying.

So we need to go deeper and stop prioritizing features. We need to start designing nudges.

The machine can tell us what users are doing, but it struggles to understand why they stop. This is where you can apply the Fogg Behavior Model. Is the user stuck because they lack the Ability (too hard) or the Motivation (don’t care)?

AI is your laborer here, not your architect. Use AI to hyper-personalize the “Ability” path—removing friction at a scale humans can’t touch. But the “Motivation”? That requires empathy. That requires a human who understands the fear, the ego, and the aspiration of the person on the other side of the screen. Your customer. The user.

2. Quantum Systems Thinking: Beyond the Linear

The old roadmap was a straight line. Feature A leads to Feature B. It was predictable, safe, and now, it’s obsolete.

AI creates a “Quantum” environment. One small tweak to a search function doesn’t just change results; it ripples through user trust, brand equity, and your long-term data flywheel. So what do we do?

AI handles the features. You handle the interconnectedness.

When you move to “Capability Ecosystems,” you aren’t asking “Does this work?” You’re asking “How does this change the essence, the soul of the product?” The machine can optimize the loop, but it cannot decide if the loop is worth building in the first place. That is a systems problem only a human can solve. Come join us at The Quantum Collective so we can learn together how be that human.

3. Blue Ocean Strategy 2.0: Outsourcing the Red

The “Red Ocean” is bloody. It’s full of competitor tracking, basic specs, price wars, and table-stakes features. It’s where most PMs spend 90% of their time. And it’s exactly what AI is best at.

Let the machine have the Red Ocean.

Use AI to automate the “commodity” work of product management. If a task is “standard,” it’s for the bot. This clears the deck for you to spend 80% of your time in the Blue Ocean.

The Blue Ocean is the “irrational” stuff. It’s the feature that shouldn’t work but does. It’s the delight that can’t be quantified. It’s the hack. The differentiation that creates a cult-like following. AI optimizes for the average; Excellence is found in the outliers. Build your Excellence habit for the AI age.

The Pivot

The marketplace is ambiguous by definition. The contributions of AI are just getting started. But ambiguity is only a threat if you’re trying to compete with the machine.

If you choose to be a Behavioral Architect, a Systems Thinker, and a Blue Ocean Strategist, you aren’t just surviving the AI era.

You’re defining it. You are creating your own job description.

Ship the uncopyable, and you will build your brand as Uncopyable PM.

The vibe in product management right now is weird.

On the surface, everything is faster. We have tools that can draft a PRD in seconds, synthesize user interviews in minutes, and generate a roadmap before you’ve finished your first coffee. We should be dancing in the streets.

Instead, there is a pervasive, low-grade hum of anxiety across the industry. A collective holding of breath.

If we are being honest, many of us feel less like conductors of innovation and more like tired crossing guards trying to manage an impossibly busy intersection built by machines.

This isn’t just about “job security.” It’s deeper. It’s existential friction. The tectonic plates of what it means to do excellent work have shifted, and we haven’t found our footing yet.

This is “The Great PM Anxiety.” It stems from three very real, very uncomfortable truths about our new reality.

1. Technological Alienation (The Empty Chair)

For a decade, our value as Product Managers was tied to our ability to translate human needs and business objectives into technical specifications. We sat at the table because we spoke both languages.

Now, a black box sits in the middle of the table. It speaks both languages faster than we ever could.

When an algorithm takes over the “how” of building, it’s easy to feel like we’ve lost agency over the “why.” This is technological alienation—the feeling that the tools are no longer serving you; you are serving the tools.

But here is the uncomfortable truth: If our sense of self-worth as a PM was tied to writing the perfect Jira epic, we were playing the wrong game all along.

The anxiety you feel isn’t the loss of your seat at the table. It’s the realization that you need to bring something entirely new to it. The machine brings syntax. We must bring soul.

2. The "Workslop" Paradox

AI promised an end to drudgery. We were promised time to think strategically.

What we got was a firehose of mediocrity.

AI is spectacularly good at generating average content at scale. I call this “Workslop.”

Suddenly, our inbox isn’t just full of human requests; it’s full of AI-generated competitor analyses, synthetic personas that feel slightly “off,” and feature ideas that are technically feasible but emotionally dead.

The cognitive load hasn’t decreased; it has skyrocketed. You used to spend energy creating. Now, you spend twice that energy filtering, verifying, and rejecting the “slop.”

Efficiency is irrelevant if we are just efficiently wading through noise. The new skill isn’t generating more ideas; it’s having the taste and the discipline to delete almost all of them.

3. The Identity Crisis

This is the hardest one.

For years, “excellence” in product management had clear markers. A perfectly groomed backlog. A pristine velocity chart. A launch that hit its date. These were verifiable proofs of our competence.

Today, an LLM can groom a backlog.

If the mechanics of the job can be outsourced to an algorithm, who are you?

We are shifting from an era of Creation to an era of Curation. This is terrifying because creation feels like work. You can point to the bricks you laid today. Curation—judgment, taste, empathy, deciding what not to build—feels fuzzy. It feels risky. How do you justify yourself in the performance review meeting?

The friction you feel in your mind is your old definition of excellence colliding with a new reality.

The Turn

The anxiety is real. But it is also a signal.

It is telling you that the era of the “Product Mechanic” is over. You cannot out-process the processors.

If you try to compete with AI on speed, volume, or syntax, you will lose. And you will be miserable while doing it.

The only viable path forward is to double down on the things the machine cannot do. It cannot feel the irrational delight of a user. It cannot navigate the complex political web of stakeholders with grace. It lacks an “inner excellence engine” to tune.

The future belongs to PMs who stop obsessing over the tools and start cultivating their own judgment.

Ship art, not slop.

Many established enterprise products are “using AI” today in roughly the same way you might use an internal combustion engine as a buggy seat-warmer.

Technically impressive. Practically absurd.

Large Language Models are being bolted onto existing products to summarize screens no one reads, generate text no one trusts, or autocomplete workflows that were never the bottleneck in the first place. The result isn’t transformation - it’s ornamentation. A powerful engine idling under a seat, producing heat where no one asked for it.

This happens because adding AI is easier than rethinking the product. Real AI leverage requires questioning product-focused assumptions: Who is this for? What problem are we actually solving? What work should disappear entirely? Those are uncomfortable questions in mature organizations built around stable processes, incremental change, and reliable cash cows. The Innovator’s Dilemma is alive and well. But the competitive wolves are at the door: somewhere some startup is re-thinking your whole solution and will soon emerge to challenge your product-market fit. Sooner than you think, probably.

Low-hanging fruit or executive appeasement?

In the meantime, teams ship “AI features.” A button. A panel. A badge. Proof that the engine exists, not that it’s being used well. Sometimes that’s justified as ‘establishing a beachhead’ or ‘low-hanging fruit’ while the really transformative work takes place elsewhere. Other times it’s done to satisfy executive pressure. These seem to be the initiatives that are failing.1

The irony is that LLMs are extraordinary tools for reshaping products - when they are treated as part of the core system, not a decorative upgrade. Used well, they collapse steps, eliminate handoffs, and change what’s economically possible. They provide the basis for re-envisioning existing products or inventing new ones that could not have existed without them.

But to do that, you don’t warm the seat.

You redesign the vehicle.

As large language models have entered mainstream consciousness, their potential to solve human problems is undeniable. For those with product-oriented worldviews, the scale of the opportunities for transformation are astonishing. That said, today’s product thinkers are trying to use this groundbreaking technology primarily to build marginal improvements and slightly different iterations of yesterday’s applications - not only ignoring truly transformative opportunities but tying themselves in knots obsessing about insoluble problems.

As part of a rigorous and disciplined research process that I just made up, I think we can break today’s mainstream AI product offerings into four classes of use cases.

1. The Search Engine

2. The Chat Thread

3. The Analyst

4. The Language Generator

For now, let’s leave the agentic case aside (5). From a product perspective, agentic AI is primarily about giving LLMs the ability to control digital systems, which makes it really just a bolt-on extension of an LLMs’ reasoning and language abstraction capabilities. Transformatively useful, yes, but really just a side feature.

Paradigm 1. The Search Engine

The general public tends to use LLMs as a search engine. Let’s be honest: it hasn’t been that long that we’ve had a single easy-to-use portal to all the world’s digital knowledge, and it is still extraordinarily valuable. Search engines have always been a proxy solution for the real human problem: a quest for answers. Google put those answers few clicks away, albeit with the added friction of having to consume the raw, unfiltered (or badly filtered) information the world had published in order to find those answers. LLMs cut out that consumption step, allowing us to leapfrog past that consumption stage and get to the answers we were looking for. From that perspective, Google was always a temporary solution mandated by the limitations of technology.

It’s no surprise, then, that the overwhelming familiarity of the general public with LLM products is about asking questions and getting answers. It’s no coincidence that the primary user experience of most public-facing LLMs looks identical to the one-click form-input paradigm search engines adopted in 1997. We would be forgiven for thinking that LLMs are just Google with fewer clicks.

and this is where a lot of the obsessive focus on disruption comes into play, because the dominance Google has in our lives—arranging all the world’s digital information and letting us access it more efficiently—creates a disruptive effect on existing industries.

Paradigm 2. The Chat Thread

The vast majority of LLM-based applications involve a chat interface, an experience almost as universal as the search engine. With fifteen years of ubiquitous text applications and live chat support experiences under our belts, multi-step chat experiences feel comfortable and accessibly human to the general public. The fact that most public-facing LLMs have chosen to start with the search engine and morph immediately into a chatbot is no surprise.

Chat is also closely related to the search engine paradigm in that a substantial share of search sessions involve iterative refinement: users submit an initial query, review results, then tweak wording or try new terms until they find a specific answer. Log-analysis and task-based research consistently show that multi-query sessions are a meaningful minority i.e., many searches are not one-and-done, they’re conversational and exploratory. Put simply: the search for answers is often a dialogue, not a one-time transaction anyway.

Chatting with an LLM feels like chatting with a human. Usually. So users already have a familiar mental model of how to do it and don’t need training or (gasp) exposure to new things.

Paradigm 3. The Analyst

In a distant third place comes what I call the analyst use case, which involves ingesting large amounts of data and finding patterns and inferences. I’m not talking about just display or arrangement of retrieved information - that fits into the search category - mean actually finding trends or connections or conclusions.

The analytical use case is happening more frequently in B2B and at an industrial scale, but even there we’ve barely begun to scratch the surface.

Paradigm 4. The Language Generator

And finally, rounding out the top four in a distant last place, is the use of LLMs to generate text. This use case is about writing copy of some kind. Thanks to the monopolies of Microsoft 365 (30-44% of the productivity software market) and Google Workspace (44-51%), the addition of Copilot and Gemini to word processing have made language generation ubiquitous and accessible, although the use of LLMs in word processing is still in its infancy. Language generation in social media is far more common (shout out to all the Russian bots on Twitter), and email/text communication (same for the vast and verdant spam farms in Southeast Asia and Los Angeles), although not technically a use case for the purposes of this story.

Into this category we can also dump literal language generation, code-generating applications like Cursor, and image and video generation like Midjourney and Google Veo - since Latvian, Javascript, and JPGs are all just different languages. This case is simultaneously the most interesting and most fraught with never-AI expressions of angst about the threat to humanity. But truthfully although they have generated a tremendous amount of ink and pixels, their impact heretofore has been minor.

But Wait There’s More … Right?

Judging from the paucity of other options in the world today, the general public would be forgiven for concluding that these four cases are synonymous with AI itself.

Clearly, this is nowhere near the end of the list. With even a modicum of imagination, it’s obvious that the product possibilities of LLMs go far beyond these four paradigms. But because we can easily imagine other ways AI can be used and yet have seen so little outside these four, suggest that we are at the earliest stage of product development. Or to use a mixed metaphor an LLM might suggest, we are just scratching the surface of the iceberg. Picking the low-hanging fruit.

These are product experiences that are easy to conceptualize, intuitive for users to understand, and straightforward to build. One can imagine significant improvements over the status quo in cost savings or efficiency, opportunities for those who figure out how to apply these paradigms in an existing industry. Sure, we may be inventing fancier buggy whips, but everyone needs a buggy whip. Right?

Data Accuracy Was Never Supposed to Be A Native Feature

One significant problem with chasing product concepts based on yesterday’s familiar paradigms is that they run straight into one of the large language model’s most consistent and annoying limitations: data accuracy. Accuracy matters greatly when one is requesting the price of a product through a chat thread. Accuracy matters greatly when one is trying to ascertain the length of a World War II battleship through a search engine. Accuracy matters greatly when you’re trying to do stock analysis or baseball statistic pattern recognition.

Unfortunately, by their very nature, language models are not designed for this kind of accuracy. They’re tailored to find inference and relationships between linguistic concepts and convey them in ways familiar to human perception. The actual numbers or facts being represented by language was never the point.

But because we’ve picked product paradigms from the early 2000s we inherit the user expectations that come along with them. A misquoted price by a chatbot is a significant bug. A hallucination about the length of a battleship is a major problem. A misdiagnosed stock price prediction can cost real money.

So we see the LLM firms chasing layered logic daisy chains, detail checking (and double-checking and triple-checking), and source citation. Downstream application builders experiment with better context and prompt engineering to make up for the weaknesses of the models. We are spending a tremendous amount of effort measuring how frequently a LLM provides the right answer and correcting what is seen as its weaknesses.

While largely ignoring their transformative strengths: linguistic abstraction, conceptual relationship association, and generative expression.

Eventually, the LLM providers are going to get better and better at this, so this doesn’t seem like a very long-term-sustainable strategy for downstream application builders. That said, there’s significant short-term opportunity in building yesterday’s products better cheaper faster, and given the cost of developing these applications dropping by a factor of twenty, we might as well get at it.

The Not-So-Low Hanging Fruit

Given all this, it might be that the real opportunities in AI product development lie beyond the product paradigms of the previous generation. Perhaps product designers should figure out how to capitalize on use cases and product concepts that play to the strengths of the models themselves, rather than fixing problems that come with the user paradigm they chose to adopt.

The range of human experience - and the problems that come with it - involves more than just data accuracy. The use of language models in counseling, coaching, and teaching - cases where accuracy is secondary to the connections between and communication of ideas - is only beginning to be explored.

That’s not to say there won’t be difficulty in getting the tone right or in avoiding bad advice, but in many aspects of the human experience, inspiration, imagination, and the connection between ideas are the areas where AI doesn’t replace humans but can provide a tremendous advantage.

Applications designed to replicate the worldview of a thought leader, creative frameworks where AI brings characters in stories to life, systems that help humans collaborate or brainstorm or imagine or interact … these are places where data ‘accuracy’ is largely irrelevant but the models’ expressive and intuitive strengths add value - making it feel interesting, valuable, wise, or useful.

That’s not to say that building these applications will be difficulty-free. Managing tone, safety, and the nuances of human expression will be challenge enough. But when done successfully, the results could be more valuable than a 2025 version of Alta Vista.

How Product Managers Suddenly Became the Most Important Job in Tech

Creating new products - or new ventures around new products - has always been risky. Ideas for problems to be solved, solution concepts, products that might help people … within a few minutes of imagination, you can come up with dozens. The question has always been: what’s the right idea? Which one of the many opportunities out there should anyone pick?

For a long time, the riskiest part of establishing a new venture or new product was the cost of marketing it. But over the past 15 years the cost of getting a product to market plummeted. Now the full-fledged embrace of digital marketing technologies has created fast and cheap ways to find customers for a given value proposition. It’s not easy, but it is cheap.

But the cost of building those products, digital ones in particular, remained stubbornly high. There was never enough engineering capacity and as a result the cost was very high.

And then came AI

The same thing that happened to marketing is now happening to development. The cost of building a new product is seeing 10× reductions … sometimes 20×, sometimes even 100×. What once cost $500,000 and six months now costs $25,000 and can be in market in 30 days.

But one thing hasn’t changed: the hard part of building products was always the thinking part: figuring out what customers needed, what solution would work, what customers want, what features to build and in what order, and rapidly adapting to customer feedback. That doesn’t change with AI. What was once regarded as a vital later-stage endeavor has suddenly become the most critical aspect of tech.

Ideas are cheap. Execution is cheap. Marketing is cheap.

Suddenly the most important question becomes: what do you do first? How do you test it? How do you adapt quickly to what the market tells you? Turns out this is what product managers have been doing all along.

Product managers spent the better part of the past 30 years quietly establishing methodologies process for deciding what to build, in what order, and how to coordinate engineering, marketing, sales, operations, and support to make it happen.

Yet it takes a while for revolutions to take hold. The world still operates with old-fashioned expectations: organizations wedded to traditional development models and expensive engineering teams still haven’t quite grappled with the new reality. Some have barely recognized that the risk curve has flipped.

The old model - write a business plan, raise money, build a product, and only then try to sell it - was expensive and risky for founders, products, and investors. Recognizing that this model is literally backwards will be the difference between products and ventures that fail and ones that succeed.

VC Is Optimized for the Wrong Century.

Until recently, new ventures and product ideas that solved problems for relatively small markets couldn’t get funded. The cost to build them required expensive people, which meant capital, which meant a certain return on investment for investors. If the size of the return didn’t offset the size of the risk, the idea simply died. The more capital required, the greater the risk, the higher the return had to be. “Go big or go home” was the only attitude that made sense.

But the world has always been full of $1M, $5M, $20M, even $50M business opportunities. They may not have unicorn-style exit potential, but they’re real, meaningful money makers and wealth generators. Yet most of these ideas haven’t fit the growth investment model of the past few decades.

Professional venture and hedge funds require significant ROI multiples, which push entrepreneurs toward solving bigger and bigger problems in hope that they might yield unicorn-level exits. With 90% failure rates, one in ten startups had to deliver HUGE returns to make the model work at all.

Everything else? Relegated to that condescending term: “lifestyle business.” If a startup couldn’t produce unicorn-level returns, it just wasn’t worth the attempt.

Then AI Happened.

Over the past 18 months, the cost to build ventures has plummeted - by 10 or twenty times. Ideas that would’ve cost a million dollars and six months to get to market now cost $50K and can be in market in 30 days.

The implications of this new risk curve are only beginning to be felt - let alone understood. Traditional investors are still wedded to the unicorn-chasing models that worked in the past: high risk, high return, extremely high failure rates. But what happens in a world where $1M can get you a $10M return, and you can do that a hundred times? What do ventures that aren’t unicorns - but elephants, buffaloes, or even gazelles - with low investment requirements do to the math?

I have personally shelved dozens of business ideas that didn’t fit the risk curve demanded by modern professional investors: too risky, too narrow a market, too unlikely to deliver the returns professional investors needed to offset huge bets and 90% failure rates.

But suddenly many of those ideas are viable: model-assisted development, rapid prototyping, cheap digital marketing experiments to find product-market-fit … the cost of getting a startup in market has dropped through the floor.

With low six figure investments, ventures can return $5M, $10M, even $20M exits … or ongoing cashflow as debt service or dividends.

The Age of Big Bets Is Over. The Future Is a Hundred Small Wins.

Anyone would tell you that if you went to Las Vegas, sat at the roulette table and placed $50M worth of chips on a single number ten times in a row - hoping that one you’d get one hit … that would be a terrible strategy. Much better would be to place lots of smaller bets with better odds: lower individual returns, perhaps, but a far higher likelihood of success in the aggregate.

Established product categories and niche markets are suddenly wide open for an AI-driven microventure approaches. Product concepts that were beyond comprehension before the invention of large language models are now weeks away from reality for those who seek them: nimble tiny teams using AI to build AI. The market is full of potential founders facing a sour job market and a world full of problems that desperately need solving.

The old unicorn model will still exist - but it will be a tiny tip of a much larger iceberg.

And nobody is doing it yet.

The role of a Product Leader is confusing.

It sits at the intersection of a dozen departments, holds endless meetings, and rarely has direct authority over anyone who builds the thing. Often, it looks like a glorified backlog manager, pushing tickets and refining requirements.

That’s the tactical job. And if that’s all you’re doing, you’ve missed the point.

The real job is not about managing the product. It’s about owning the uncertainty. It’s about drawing a line in the sand where confusion meets conviction.

The Product Leader is The Editor.

Every company has infinite things it could build. And every customer has infinite demands.

The Product Leader doesn’t simply collect requests. They are the chief editor, armed with a giant, heavy pair of shears.

The art isn’t in what you include - it’s in the courage to delete what is merely good to make room for what is great. And to make room to what is essential for your minimum viable market. The job is to make a ruthless, profitable, and honest choice about where the organization’s energy must be focused. If you aren’t saying “no” far more than you say “yes,” you’re not leading. You’re compiling.

Product vs. The Others.

When everyone is running in the same direction, where does the Product Leader stand? They stand at the map table, with a compass and a telescope, while everyone else has a hammer, a megaphone, or a fire extinguisher.

Versus The CEO (The Vision).

The CEO sets the destination—We need to conquer the North Pole.

The Product Leader draws the map and estimates the time of arrival—To get to the North Pole, we must first secure the sled dogs, stock the right supplies, and avoid the treacherous iceberg at this specific coordinate.

The CEO owns the ultimate success. The Product Leader owns the strategic path.

Versus Engineering (The How).

Engineering is the brilliant chef in the kitchen. They decide on the best techniques, the sharpest knives, and the most efficient cooking method. They guarantee the food is technically perfect.

The Product Leader is the restaurateur who decided the restaurant should be French, not Italian, because the neighborhood desperately needs a great French bistro.

The Engineer owns the quality of the build. The Product Leader owns the quality of the market fit.

Versus Marketing (The Storyteller).

Marketing is the brilliant publicist, writing the compelling copy and placing the ads that get people to the door.

But the Product Leader wrote the menu. And before that, they tasted the food and decided if the dish was good enough to be served in the first place.

Marketing tells the story of the product. The Product Leader is the story.

Versus Customer Support (The Firefighter).

Support is the indispensable firefighter, racing to put out the blaze that is happening right now. Their job is immediate resolution, empathy, and damage control.

The Product Leader’s job is to look at the patterns of the fires and rewrite the building code so they never happen again.

Support solves the current problem. Product prevents the next system failure.

The True Power: Constraint.

Your power doesn’t come from having the largest team or the biggest budget. It comes from the ability to impose constraint.

Constraint clarifies. Constraint forces focus. Constraint is the foundation of innovation. When you confidently state why you are building something—the specific market problem, the non-negotiable solution—you give everyone else the boundary they need to excel.

A company with a Product Leader is a ship with a clear rudder. A company without one is a thousand smart people rowing in a circle.

Choose to be the rudder.

• Product gives more. More market reach, more customers, more loyalty.

• Product is driven by better. Better experience, better community, better outcomes.

• Product creates culture. Status, feel, belonging to your tribe, identity.

• Most of all, product creates change.

• Change the product, change the culture, change the world.

• The Product leader makes change happen.

The root and the tree

How tall is the tree? This is what most people care about. The taller the tree, the bigger shadow it casts. The more shadow (benefit), the more you can profit from it. Too many businesses spend their time running a features-and-benefits show, trying to get the most out of the shadow.

The thing is, tall trees need deep and strong root systems. Without them they fall with the first storm. The whole business often follows.

This is a book about growing deep roots. About anchoring your product deeply in the minds and hearts of those you seek to serve. And it's about creating products that drive the market, not simply being market-driven. Product is about doing work that matters for the people who care.

Product isn't just marketing and sales

Nor is it about strategy and metrics. You built an amazing product. Your boss wants more sales faster. The business needs more cash flow. Operations are looking for shortcuts. The marketing department wants to build on the brand equity. And the sales people want to meet their quarterly quota. How do you balance all these priorities?

Then you remember you serve the customer. You create the product, you plant the tree and water it. Because you know that the customer wants to use of its shadow.

This is why Product is not just of strategic importance. It is central to any business. You have a Product question and it's possible that the market has an answer. But only if you look for it.

Seth Godin said that the best way to complain is to make things better. And the first step on the path to making things better is to make better things.

And better happens when the market embraces what we are offering. Product makes that happen. This is a book about the Product people.