The mirage of the "AI Product Manager"

Before we get into it, a quick disclaimer: this is just my opinion based on my experiences, so before you grab your pitchforks, read it first.

There is an anxious, high-velocity effort across the tech industry right now to summon “AI Product Managers (PM)” into existence. If you look at a modern tech job board, you would think human civilisation had discovered a new element. Org charts are being torn apart to accommodate the title. Thousands of people are frantically updating their LinkedIn profiles, driven by a collective panic that if they don’t rebrand by COP Friday, they’ll become obsolete.

But if you stop and ask what the title actually means, the edges blur immediately. To some, it means they use Midjourney to mock up a UI. To others, it means they are wrapping a third-party API into a standard product loop and calling it a feature. For a lot of people, they aren’t entirely sure what it means at all. They just know they don’t want to be left behind.

The underlying assumption seems to be that artificial intelligence is a standalone product vertical. Sacred, novel, requiring its own breed of PM to manage the deployment of the models themselves. What I think is actually happening is that we’ve mistaken a tool for a strategy. And that’s not a new mistake. It’s just a particularly expensive one to make right now.

And look, I get it, up to a point. The tooling is genuinely different. The stakes in certain domains are genuinely different. But there’s something about the pace at which this title is proliferating that feels less like a disciplined response to real specialisation, and more like the classic institutional move: reorganise hard enough, and the strategy will sort itself out.

The job title is doing something though.

When a company creates an “AI PM” role, they’ve already made a structural decision: that AI is the thing being managed, rather than one of several paths to solving something. The job title is the symptom. The underlying condition is that somewhere upstream, in a strategy offsite that probably had a lot of Post-it notes, someone decided that AI was a product category rather than a capability. And now the org chart reflects that decision, and the decision feels very hard to undo.

Despite the earned eyebrow raise you’re probably giving me right now, this isn’t me being dismissive of AI. Quite the opposite. I think AI matters enough that it’s already beginning the long process of disappearing into infrastructure. 

Here’s the thing about capabilities that become genuinely ubiquitous: the job title built around them stops making sense. AI may become ubiquitous enough that being an “AI PM” eventually sounds as vague as calling someone an “internet PM.” Eventually, the technology becomes infrastructure, and the title quietly dissolves because, of course, you’re accounting for it. That’s just the environment now.

I want to be careful here because there’s a version of this argument that’s just smug, and I’m not trying to write that. I’m not going to tell you AI isn’t important or that the roles don’t exist. Some of them absolutely do. There are real, specialised roles emerging in the AI space that deserve their own category: AI governance and safety, foundational models, inference infrastructure, evaluation systems, agents, etc. These require a depth of fluency that’s meaningfully different from general PM scope. But outside of that specific layer? A lot of what’s being labelled “AI Product Management” is just… product management.

Every major platform shift follows the same arc. First the technology becomes the organising principle. Then it becomes infrastructure. Then differentiation moves upward: into judgement, orchestration, governance, strategic clarity. The internet, mobile, cloud. All of them. There’s no particular reason to think AI is exempt from that pattern. If anything, the fact it rarely operates in isolation makes the eventual shift more significant, not less.

Google I/O this week was actually a useful illustration of this. Once you looked past the keynote polish, most of what shipped weren’t framed as “AI products”, but rather existing systems becoming more capable: Search, Maps, developer tooling, consumer workflows. AI was one implementation layer among many. The framing remained the human problem. Here is the friction that real people experience. Here are the tools we’re using to reduce it. AI is one of those tools. So is fifteen years of Maps infrastructure sitting underneath it.\

And I think that’s the point. Not because Google is doing it right and everyone else is doing it wrong, but because it’s a useful illustration of what platform product thinking looks like when the tool doesn’t eat the strategy. The companies I find most interesting right now aren’t asking “how do we build AI products?” They’re asking “what is the actual friction our users are experiencing, and which tools are the right path to reducing it?” The AI ends up in the solution. It doesn’t start there.

So if execution gets commoditised, if writing the loop, connecting the endpoint, deploying the feature are increasingly things that just happen, what doesn’t get commoditised? Judgement, probably. The ability to look at a human problem, understand the constraints around it, and decide what to actually build. Not what’s possible. Not what’s impressive to demo. What’s worth building, for whom, under what conditions, and what not to build at all. 

I wonder if part of the “AI PM” proliferation is just what happens when role taxonomy gets ahead of strategy. The quiet institutional hope that if you summon enough of the right-sounding titles, coherent product direction will emerge automatically. It usually doesn’t. It tends to add a layer of coordination overhead between the human problem and the people trying to solve it.

The organisations that actually get this right probably aren’t the ones with the longest list of AI PMs; they’re the ones that stayed focused on the human problem long enough to figure out which tool, if any, was actually worth reaching for. Which is, and always has been, just the job. My favourite part of the job!