20 Lessons I Learned in 20 Years as a Product Manager

This year marks my 20th year in Product Management! Yay!

Twenty years is long enough to see the product industry reinvent the language of the job several times.

New frameworks. New org structures. New tools. New rituals. We are now entering another major shift in the world of AI-powered products and AI-PMs.

Agentic systems will change how product teams operate. Teams will increasingly build internal operating systems composed of multiple agents to generate ideas, run validation loops, prototype solutions, and accelerate execution.

Every PM will likely have a "token budget" assigned in their salary package. And the cost of testing and building will fall dramatically. That is great news for innovation.

While we might see that everything is changing faster than we can keep up:

• The mindset that gives direction to the agents still matters. The perspective that orchestrates the system still matters.

• The judgment that decides what to build, why it matters, and what good looks like still matters.

• And the accountability for the outcome still sits with people

At its core, Product Management is still about seeing clearly, deciding well, and helping teams build something that matters.

If you are trying to understand the craft of Product Management, here are 20 lessons I have learnt that can help you navigate through this AI era and your PM journey:

1. Product Management is not a single skill. It is a bundle of random skills.

Every thought leader is describing one side of the elephant. In some ways, they have to, because they need a niche to teach.

The AI shift makes this even more obvious. As agents take on more research, writing, prototyping, and validation work, it becomes even easier to mistake one capability for the whole job.

AI will give every employee leverage for the cost of a token. But it does not collapse the craft into prompt writing. Product Management has never been one thing. It is still a bundle of random disciplines:

• Customer understanding

• Business judgment

• Prioritisation

• Systems thinking

• Taste

• Communication

• And most important of all, accountability of outcome

The real work is still building your own point of view by learning from many perspectives, not blindly following one tool, one guru, or one framework.

2. Market quality shapes more than people think.

Quality of competition is correlated to the size of the market.

The bigger the market, the more positive the sum of the competition is. The smaller the market, the more zero-sum the competition is.

Good operators in weak markets often look worse than they are. Great markets pull the products out of teams. The market you swim in shapes what is possible.   

AI does not remove that truth. It amplifies it. When the cost of generating ideas, prototypes, and tests falls, more teams can build faster.

But speed does not create demand.

A weak market with faster tools is still a weak market. A strong market with faster tools becomes even more fertile. That is why PMs need to get better at distinguishing real market pull from tool-fuelled output. AI can help teams create more. It cannot make the pond bigger.

3. Progress beats perfection, but details still matter.

Always strike a fine line between being a perfectionist and a progressionist.

Sweating details matters. But so does momentum.

Great PMs know when to refine and when to move. Be a perfectionist in details and a progressionist in the vision.   

This tension matters even more in the age of AI.

Agentic workflows will dramatically reduce the cost of building drafts, experiments, mockups, and prototypes. That is good.

But it also creates a new trap: confusing faster output with better judgment.

AI makes it easier to move. It does not decide where rigour matters most. Details still shape trust, coherence, and quality. Vision still determines where speed should be applied.

The craft is still in knowing what must be right now, and what can improve in motion.

4. No amount of research, or AI, replaces walking in the customer’s shoes.

Research matters. AI matters. But they have limits.

At some point you have to get close enough to feel the problem. To see the friction. To feel the pain. To understand the context people are actually operating in.

Without context, there is no understanding. Without understanding, there is no product. 

This may be one of the most important product lessons in an AI-native world. AI can help summarise interviews, cluster patterns, surface opportunities, and compress huge amounts of information.

But that is still an inferred understanding. It is not lived understanding.

Customer pain is not just a data point. It is something experienced in context. If you do not get close enough to real people and real situations, you end up designing for generality - an extremely high-level product that we see as "AI slop". AI can compute reality. It cannot replace contact with it.

5. Knowledge lights the path. Execution leads the way.

A lot of people learn products like spectators. Courses. Frameworks. Certifications. Content.

Most advice you hear fails on execution, not because it is bad, but because it needs to be adapted.

You only do real learning when you make decisions, do real work, fail in public, reflect honestly, and do it again.

Doing it again is the unique insight that no one has. That is what makes you valuable. 

AI can compress the spectator phase. It can help people sound informed much faster. That is useful, but dangerous too.

Knowing more, faster, still does not equal capability. Execution remains where truth appears. Execution is where political friction, technical constraints, messy stakeholders, and real-world trade-offs show up.

AI can support execution. It cannot replace the learning that comes from being accountable for the outcome.

6. Copy fully, or not at all.

Shuhari is how Japanese mastery works: first copy the form from the master, then detach from the master, then create your own.

If you skip the PM fundamentals, you cannot adapt to shifts in the market. If you cannot adapt, you cannot become exceptional. Mastery is not random originality.

It is deep understanding first, then thoughtful evolution.

This applies directly to PM in the AI world. Right now, many teams are copying the surface of AI maturity: prompts, agent setups, copilots, eval loops, dashboards, and orchestration patterns.

But if they do not understand the fundamentals beneath them, they will break the moment the context shifts. Copying is fine. Half-copying is not. Learn the principles, not just the artefacts. Then evolve them. That is how you move from imitation to mastery.

7. Judge yourself based on your decision system, not the outcome.

Bad decisions are expensive. You pay for it with wrong bets, bad launches, weak hires, and features nobody wanted.

Judge yourself based on your decision system rather than the outcome. You cannot control the outcome, but you can always control your system.

This thinking stops you fearing mistakes and allows you to learn and optimise decisions. 

In an AI-enabled world, when teams can generate more options faster, it becomes easier to mistake activity for intelligence.

A polished AI output can create false confidence. That is why your decision system matters more than ever. What assumptions did you use? What evidence mattered? What trade-offs did you make? What signal did you ignore?

AI can give you more input. It is still your job to ensure the system produces sound decisions.

8. Strong product work finds the win-win.

Great products do not just delight customers. They also work for the business.

Product is the bridge between customer value and business value. Need both to succeed.

If it solves a customer problem but creates no business value, it will not sustain. If it only serves the business and creates no customer value, it becomes extractive. 

That is why the PM’s role still matters. AI expands what can be built. It cannot take responsibility for ensuring that the balance between business and customer value is mutual - that is the PMs job.

9. Strategy is choice, and every choice has a reverse effect.

Every yes creates a no somewhere else. Every path chosen closes other paths.

That is why strategy is not a list of everything you want to do.

Strategy is deciding what matters enough to focus on, and having the courage to say no to million-dollar ideas to chase billion-dollar ideas. 

AI will tempt teams to forget this. When build costs drop, everything starts to look doable. More ideas feel within reach. More experiments feel cheap. But strategy does not become less important when building gets easier. It becomes more important.

Lower execution cost increases the risk of scattered focus. The discipline is still deciding what deserves attention, even when the buffet is all-you-can-eat, which does not mean you should consume everything.

10. Be a farmer, or a hunter. Not both.

Hunters chase immediate wins. Farmers cultivate long-term value.

The product you work on determines which game you are playing.

Customer-facing products often require hunter PMs. Platform products often require farmer PMs.

Hunt when speed matters. Farm when compounding matters. But do not confuse one for the other, as it will only lead to frustration.

AI increases leverage for both types. Hunters can run faster loops. Farmers can build better systems. But it does not remove the distinction.

A customer product optimised for short-cycle learning is still a different game from a platform product optimised for long-term compounding. The mistake is to use AI speed to apply hunter logic everywhere. Some problems still need cultivation, not chasing.

11. Don’t let the horse lead the coachman.

Metrics are indicators, not the objective. The objective leads. The metric follows. The moment a team starts chasing the metric instead of the mission, judgment gets distorted.   

I've made a lot of mistakes in my days by focusing on one metric myopically and losing track of strategy shifts, market changes, customer shifts, and platform shifts.

Metrics are great to gauge the direction, like a compass, but it does not tell you where to go, like a map. (That's an objective or a goal)

Nowadays, we call this combination evaluations for an AI agent. Agents can optimise aggressively against whatever signal you give them. If the goals are weak, the system scales the wrong behaviour faster than humans ever could. That's why evals are important because it helps you set the right direction for your agents

That is why the coachman still has to lead. The objective is the anchor. Metrics are instruments, not masters. AI can optimise. Only humans can decide what is worth optimising.

12. PM is not a ceremony job. It is a decision-making job.

The job is not to say no all day. The job is to make high-quality decisions that create business value.

That means having a point of view. A point of view emerges from market, customer, and business signals.

Choose which signal matters, ignore the rest, and be accountable for the trade-offs.   

This is one of the biggest reasons AI will not replace PMs.

It may reduce PM ceremony. It may draft docs, cluster insights, and prepare an analysis that strips away some admin. But what remains is the real job: deciding well. The role was never meant to be about process theatre. It was always about judgment. AI makes that more visible, not less.

13. Tech change. Product mindset lasts.

Every era pushes a new set of tools. 20 years ago, it was mobile. Then social. Then cloud. Now AI. AI augments intelligence. It does not replace it. 

Strong PMs learn the tech and master them, but do not worship them.

Tech help you build. Fundamentals help you think. Process helps you scale. PMs who last do not confuse the tool with the job. Solve the right problem, and the tech will come. 

The PMs who endure are not the ones who chase every tool wave with religious fervour. They are the ones who can absorb new tools while keeping the deeper craft intact.

In the future, PM teams may operate more like agencies of agents than managees of mid-level individual contributors. Even then, the mindset that gives those agents high-level direction remains scarce.

14. Politics clouds good judgment and kills good products.

Good products die all the time because of ego, turf, and status games. The moment internal politics starts outweighing customer truth, the product suffers.

You cannot eliminate politics. But you can keep coming back to one question: what is actually best for the customer and the business?   

AI does not remove politics. In some ways, it gives politics better camouflage. Weak thinking can now be wrapped in more polished language, faster analysis, and a stronger illusion of certainty.

That makes judgment even more important. The core discipline is still the same: bring the room back to reality, not just to whatever Claude generated that sounds convincing.

15. Missionaries always beat mercenaries in the long run.

Mercenaries work for the pay. Missionaries work for the mission.

The best teams care deeply. They push harder. They stay longer. They build with conviction.

In product, mission matters. People can feel when a team truly cares.   

When AI tools make output easier, conviction becomes a stronger differentiator.

Teams that care will still ask the harder questions, defend quality, stay with ambiguity longer, and push through the messy middle.

AI can expand capacity. It cannot manufacture human will. (Yet...)

16. Competitor awareness matters. But obsession kills originality.

The moment you anchor too hard to competitors, you start building from their frame.

The companies that create new categories do not just benchmark. They absorb what is useful, then build from their own conviction.

Comparison is the thief of innovation. Do not compare so much that you lose your point of view.   

Now it is even easier to take a screenshot of a website and ask any AI agent to code it up for you in 8 seconds. It is simple to use Deep research to scrape, summarise, benchmark, and react. But being better informed does not guarantee originality.

In fact, constant comparison can make teams more derivative. For example, why do all AI tools have a chat interface? AI can help you see the field more clearly. It is still your job not to become trapped in someone else’s frame.

17. Strong product craft is often subtractive.

The first stage of technology is analogy. The last stage is when it becomes invisible. The middle is where we risk capital.

Complexity is easy to add and hard to remove. Simplicity is hard to create and easy to underestimate.

Before you build something new, ask whether the smarter move is to remove something old. 

AI makes addition dramatically easier. That is exactly why subtraction becomes more valuable.

If agents can generate more features, flows, surfaces, assistants, and layers of complexity, then product craft must become even more disciplined about what not to add.

The end state of strong technology is not maximal visibility. Its usefulness is so natural it fades into the background. Invisible, trusted, simple. That is still the bar.

18. Law of 80/20: there are more bad managers than good ones.

Most people will work for managers who are average at best, and poor at worst.

So one of the most useful skills in a career is learning how to work around weak management and protect your reputation.

A bad manager is not just a frustration. They are a warning about the leader you could become if you are not careful.   

AI will not fix bad leadership. Leadership is still about self-awareness and knowing others. It's still about judgment, trust, clarity, and creating conditions for people to do good work.

The danger is confusing tooling with leadership. Tooling just helps with management. Leadership is a mindset and a perspective If anything, better tools is what will fork the bad leaders from the great ones.

19. It’s the less-travelled path where hidden value lives.

Crowded markets feel safer because other people are already there. But if everyone can see the opportunity, it is usually smaller than it looks.

Real upside often sits where fewer people are willing to go. Go there, and bring back the gold.   

Many teams will chase the same visible opportunities, the same AI-agent use cases, the same product patterns, and the same language.

That makes original insight even more valuable. The point is not to be contrarian for the theatre. It is to recognise that real upside often lives where fewer people are willing to think deeply, commit early, and build in a growing market with conviction.

20. The real job is to make a difference, not just to earn a salary.

I’m not in the job just to ship tickets, climb titles, or collect compensation. I’m trying to build something useful, meaningful, and lasting.

It's not about making money and retire. It's about making a difference and inspire.

After 20 years, I think this lesson matters more to me now than ever.   

This feels even more important now because AI will increase what is possible. But more possibilities do not automatically mean more meaning.

The fact that we can build more, faster, does not answer what deserves to exist.

That is still a human question, a Product question. It is not just about efficient construction. It is about efficient construction -> with intention.

Closing thought

After 20 years, I think Product Management is still one of the best jobs in the world.

Not because it is glamorous. Not because it gives you control. And definitely not because it comes with easy answers.

It is one of the best jobs because it forces you to think clearly, choose wisely, and build something real in the middle of constraints, trade-offs, and human messiness.

The job can be summarised into 3 words:

Build what matters.

AI will help teams generate ideas faster, validate faster, prototype faster, and execute faster. That is a good thing. It will expand what small teams can do and lower the cost of innovation.

But the mindset that gives the agents direction has not changed. The perspective that directs the agents has not changed. The judgment required to decide what matters has not changed. And the accountability for the outcome has not changed.

That was true before AI. It is true with AI. And I hope it will still be true for the next 20 years after this wave matures.

Your call to action:

These 20 lessons are the foundation of how I teach Product Management.

If you want to skip a few years of figuring it out the hard way, I’m running a course on Maven where I break down the craft, the thinking, and help you build a real product with AI in production.

My Newsletter subscribers will get 50% off the price, cohort is kicking off Next Week!

Check it out here → https://maven.com/productacademy/master-product-craft-with-ai?promoCode=PA50