AI Killed Product Management

Why Institutional Product Management No Longer Fits the World We’re Entering

The title is intentionally provocative, but the argument itself is not an attack on care, rigor, or thinking. Quite the opposite. This is an argument for taste, creativity, and playfulness, and against a form of Product Management that gradually replaced those qualities with process, abstraction, and overthinking. What is breaking down is not product thinking, but the institutional harness that shackled it.

Product Management emerged in a very specific world. Software was expensive to build, slow to change, and difficult to undo. Decisions carried real weight because mistakes lingered. In that environment, thinking ahead was not a luxury but a necessity. The role existed to protect teams from costly missteps, to slow things down just enough to avoid irreversible damage. Product Management was not originally about vision or taste. It was about protection. It functioned as an insurance policy against scarcity.

Over time, that pragmatism hardened into a discipline of its own. Product Management separated itself from engineering, developed its own abstractions, and began to observe the act of building from a distance. What started as a practical layer of risk mitigation slowly turned into a system of control. Roadmaps stretched further into the future. Discovery phases outgrew delivery phases. Tools multiplied. Process solidified. The original intent faded, replaced by the belief that uncertainty could be managed away if only enough thinking happened upfront.

The result was not better products, but heavier ones. Products shaped by alignment rather than conviction, optimized for internal consensus instead of human resonance. Speed suffered. Responsibility blurred. Creativity became something to be justified rather than exercised. The discipline designed to reduce risk quietly became one of the biggest risks to meaningful progress.

What makes the current moment different is not a philosophical shift, but a technological one. AI fundamentally alters the economics of software creation. The cost of building has collapsed, not only in code generation, but across design, research, writing, testing, and iteration. What once required teams and months can now be explored by individuals in days. Prototypes are no longer expensive commitments. They are disposable probes into reality.

This change strikes at the very foundation institutional Product Management was built to protect.

When experimentation is cheap and fast, the premise of heavy upfront planning begins to break down. You no longer need elaborate systems to predict what users might want. You can put something real in front of them and observe what happens. The fastest path to truth is no longer alignment or discovery frameworks, but exposure. AI accelerates this loop to the point where many traditional Product Management tools feel not just slow, but fundamentally misaligned with how learning now happens.

At the same time, AI enables a level of individualisation that mass-product thinking never handled well. Product Management evolved in a world of averages, personas, and segments. AI pushes us toward software that adapts to individuals rather than categories. When products can shape themselves around a single user, long-term roadmaps and fixed feature sets lose much of their relevance. Direction, taste, and responsiveness begin to matter more than prediction.

This is where my own skepticism enters. I have yet to experience a product that truly moved me because of Product Management. The products that stayed with me did so because someone had a point of view. Because someone cared deeply about how something felt, not just how it performed on a dashboard. Because someone was willing to decide rather than facilitate. That does not mean Product Management was useless. It means it was never the source of what made products meaningful in the first place.

What is quietly returning alongside these shifts is something we largely pushed out of professional software: play.

For a long time, “just building” was treated as irresponsible. Playfulness belonged in side projects and weekends, not in serious product work. In a world where software was expensive and slow to undo, that attitude was rational. Curiosity without permission was a liability. Enjoyment needed justification.

Yet many of the systems we still admire were born precisely in that space. Doom is often mentioned not as nostalgia, but as a reminder of what becomes possible when constraints align. Its early development at id Software was carried by a very small core team where individuals held clear ownership over large, meaningful parts of the system. They were not executing a roadmap produced at a distance. They were focused, directly responsible, and driven by curiosity and the joy of making something new. The point is not romance or hero worship. The point is that when overhead recedes and responsibility concentrates, craft has room to surface.

We see echoes of this dynamic again today. Experiments like OpenClaw or early agent systems such as Pi Agent did not emerge from mature product organizations or carefully staged discovery processes. They emerged because curious, capable people wanted to see what would happen if they pushed the tools a little further. In the case of OpenClaw, the adoption and emergent behavior are fascinating, and at the same time they pose serious risks in terms of security and unknown side effects. Those concerns are real and should not be dismissed. Even so, I believe the good of this kind of exploration outweighs the bad. It is precisely through such curiosity-driven experiments that new paradigms become visible.

This matters because it reveals the true driver behind these projects. It is not recklessness for its own sake, and it is not a rejection of responsibility. It is a shift in economics. When the cost of building drops far enough, the barrier to exploration collapses with it. Financial freedom helps, but it is the reduction in build cost that turns this mode of work from an exception into something repeatable and widespread.

AI changes the risk profile of play. When iteration is fast and reversal is cheap, play becomes informative rather than wasteful. You can build something extreme, observe where it breaks, and learn faster than any abstract planning process would allow. The learning comes from contact with reality, not from prediction.

This is also where the argument connects directly to the end of SaaS as we know it. When individuals can build software that works for them at low cost and with full ownership, the logic of centralized products begins to unravel. The same forces that make individual software viable also make playful exploration viable. Both are symptoms of the same shift. Creation has become cheap enough that justification no longer needs to come first.

Product Management, as an institution, was built to manage scarcity. Play thrives in abundance. That is the tension we are now feeling.

This does not mean judgment disappears. It means it moves closer to the act of building. Responsibility shifts from process to people. Taste matters again. Curiosity matters again. Enjoyment is no longer a red flag, but often a signal that something worth pursuing is happening.

Of course, large organizations will continue to employ product managers. At scale, coordination and translation remain unavoidable. But scale explains persistence, not inevitability. The fact that something exists everywhere does not mean it represents the future. More often, it represents a compromise with complexity.

What is being dismantled now is not a role, but a distance. The distance between thinking and building. Between deciding and doing. AI collapses that distance. It rewards people who can hold intent and execution together, who can move from idea to artifact without layers in between. In that world, the valuable parts of Product Management survive, but stripped of bureaucracy. Taste survives. Direction survives. Judgment survives. What disappears is the illusion that control can substitute for contact with reality.

Product Management is not ending because people failed at it. It is ending because the conditions that justified it no longer hold. Control made sense when change was expensive. Today, curiosity scales better. Iteration outperforms prediction. Learning outpaces planning.

This is not a loss. It is an opening.

It means we can build again with immediacy and honesty. We can put things into the world, observe what happens, and respond without ceremony. We can create software that is personal, adaptive, and alive, rather than broadly acceptable and static.

Seen through that lens, AI did not kill product thinking. It killed the over-institutionalized version of Product Management that mistook overthinking for wisdom.

And I think that is something worth celebrating.