A generation of product people is about to discover that they were trained for the wrong job.
Most of them are smart, structured and good at learning the game they are given, and for a long time that was enough. The game was to learn the customer, write the PRD, prioritise the roadmap, manage engineering capacity, align the stakeholders, say no elegantly, and ship the least embarrassing version of whatever the organisation could agree on. That was called product management, and in the next era much of it will be called project management.
The difference is taste.
The word is used lazily, for “make it beautiful”, for Apple worship and minimalist typography, for a founder saying “this feels premium” while pointing at rounded corners. Taste is harder than that. Taste is the invisible memory of paying attention. It is what remains after you have studied the market, the economics, the user, the architecture, the procurement reality, the power map and the incentives, and then forgotten all of it for long enough to ask the one question that matters: what will people respond to?
The answer is rarely what they will defend in a meeting, what the expert said or what the loudest stakeholder asked for; it is what makes them lean forward and say “I want that” before they can explain why. That is product, and everything else is administration around product.
For twenty years software was expensive enough that administration could masquerade as judgement. Saying no was often correct, because saying yes meant a team, a sprint, a QA cycle, documentation, maintenance, integration debt, and some poor product manager turning a simple request into a quarterly planning argument. So we trained ourselves to ask a reasonable question: is it worth it? But when the cost of trying collapses, that question starts lying to you.
A year ago, if a user asked for a Word version of something, the reasonable answer was probably no: too much scope, too much surface area, too much of the wrong kind of software. Today that instinct may be wrong. Word has not become exciting, and it is still mostly where enterprise intent goes to die. But if users respond emotionally to having a Word version, and building one costs a fraction of what it did two years ago, the product answer stops being “we are a platform, we don’t do that” and becomes a question: why do they want it so badly?
Word might stand for control, or ownership, or the ability to show the work to someone else without explaining a new system, or the comfort of an artefact they know how to judge. The job is neither to obey the user literally nor to dismiss them because the request sounds unsophisticated, but to understand the emotional primitive underneath the request and package it in a way that makes the future acceptable. Most users cannot tell you what sexy means. You still need to know.
This transition is dangerous for product managers who built their identity on being good students of the old economics of software. Experience matters, but in a platform shift it has as much downside as upside, and what made you sensible yesterday can make you blind tomorrow. BlackBerry never forgot how to make phones. It could understand the iPhone only as the product it already knew, a BlackBerry without a keyboard, and it missed that the iPhone was a computer in your pocket.
A lot of software people are making that mistake with AI. They see copilots, agents and buttons being clicked faster, which is yesterday’s software with a chatbot glued to it, and they do not yet feel that the economics of making software have changed. Until you feel it, you cannot use it.
For product managers the job is becoming harder, because AI removes the excuses that protected mediocre product thinking. You can no longer hide behind engineering capacity, say “too expensive” by reflex, or treat engineers as the sole source of truth on what is possible. Engineers are having their Kodak moment too. They know what was hard, what was expensive and what the architecture used to punish, and the future does not care what used to be hard.
The best product managers will learn enough agentic coding and systems thinking to push back, not because every PM should become an engineer, but because you cannot manage the future of software without a feel for what can now be made. Reading about it is not enough. You have to build something stupid, fun and yours, whether that is a Warhammer simulator, a Magic: The Gathering tool, a fake product for a real obsession, or anything else that keeps you up too late.
Do not build it for work. Work has too much status attached to it, and status makes you perform competence instead of learning. Build something playful enough that you can be wrong without defending yourself, because that is where taste starts waking up again.
Taste cannot be taught as a framework, because it is the scar tissue of judgement. It comes from making things, watching people react, hating the wrong thing, being embarrassed, changing your mind, trying again, and slowly learning which of your instincts deserve trust.
Our education system destroys this. School teaches you to look outside yourself for authority: the teacher tells you whether it is good, the grade tells you whether you were right, and the institution tells you whether you are qualified. Companies continue the training with performance reviews, benchmarks, best practices, the committee and the senior person in the room. Most people spend their first twenty-five years being rewarded for guessing what an authority figure wants, and then enter product roles where the job is to know what the future wants before any authority can validate it. No wonder people freeze.
No one is coming to tell you that your taste is correct. If it is any good, a lot of people will tell you at first that it is wrong, and that is the price. Taste requires doing what other people do not yet accept, looking unreasonable for a while, and being snobby about the wrong choice, because most bad products are ruined not by one catastrophic mistake but by a thousand reasonable compromises. The button is fine, the onboarding is fine, the language is fine, the feature is fine, and then one day the product is dead and everyone is surprised.
Taste is the refusal to let “fine” accumulate.
A British investor once visited a friend in France and observed that his house was bigger. Bigger was the entire comment, and it is the most Anglo-Saxon mistake there is. No one in France with any taste would pick bigger over better, where better means proportion, light, the right chair in the right corner, the old table that makes no sense on paper but holds the room together, and the way the kitchen feels at 7pm when people are opening wine and arguing about nothing.
Ask the French person how they know, and it becomes hard to explain. They just know, and there is no magic in it, because they have lived inside accumulated discernment: meals, rooms, streets, clothes, rituals, mistakes, arguments, standards. Enough exposure creates judgement, enough judgement becomes instinct, and enough instinct becomes taste. Taste is compressed experience.
Business culture trains this out of people by replacing discernment with metrics, context with dashboards and judgement with social proof. The metrics matter, and great product people usually know the numbers better than anyone else in the room, but the final act of product is synthesis, not analysis. You learn everything, and then you return to the primary ingredients. What does this person fear, and what do they want to feel in control of? What would make them trust us, make them proud to show this to someone else, and make the product disappear into the work?
The best taste becomes invisible. No one praises the thousand decisions that made something feel obvious; they use it, they trust it, and they cannot imagine it any other way. Those decisions are the last ten per cent of magic, and they are what product managers should be fighting for, ahead of more features, more roadmap or more process.
With AI that ten per cent becomes the whole game. Workflow is moving out of vertical SaaS and into horizontal layers, agentic systems, workblocks, generated interfaces and whatever comes next, and the old application boundaries matter less when software can be created around the problem instead of forcing the problem into the software. “We have the workflow” stops being an answer. What is left is product: the ability to understand a category so deeply that you can create what people did not know how to ask for, and have it feel like the future.
This is why the best product managers will become far more valuable, and everyone else will become a coordinator of other people’s decisions. That is harsh, and it is also liberating, because the path is open. Study like you are back at university, without the comforting lie of grades. Build like a teenager with no reputation to protect. Research until you understand the domain better than the experts who know only the process. Learn the new tools until the economics stop being something you can explain and become something you feel in your hands.
Then trust yourself enough to fight, most of all when you are wrong, because being wrong is part of how taste forms. You do not get taste by being correct in public; you get it by making enough real things that reality starts giving you feedback directly. That feedback will not always be polite. Run towards it.
That takes more courage than the old job did. The past rewarded people who could navigate the organisation, and the future will reward people who can create conviction where the organisation has none. It is a harder job and a much better one, because for the first time in a long time product management can stop being the art of managing constraints and become the art of creating again.
Software is getting cheaper. Taste is getting more expensive. The people who learn to taste again will build what the rest of the world understands only after it feels obvious.