Product Management Isn't Dying. It's Gone Full Stack.

Every PM I know has read Nikhyl Singhal's line about half the product management population being at risk over the next two years, and felt something twist in their stomach.

Nikhyl Singhal, former product exec at Meta, Google, and Credit Karma, went deep on this with Lenny Rachitsky: why the next two years will likely be the most chaotic stretch product management has seen, and his prediction that some companies will shed tens of thousands of roles and rehire a fraction of that number, AI-first. Listen here.

I have watched this movie before. "PM is dying" gets said every three or four years: outsourced, automated, replaced by engineering, replaced by growth, replaced by data.

It never quite happened. But I do not think the "half" framing is the useful part of what Singhal is saying. The useful part is narrower.

If your job sits anywhere in the chain that turns a business idea into shipped software, that chain just got rewired, and you need to know what you bring to it now.

📞 The telephone problemLink to this section

Here is the chain: business need becomes a spec. Spec becomes UX. UX becomes development. Development becomes what actually ships.

Every one of those handoffs used to cost you something. Not maliciously, just entropy.

By the time something shipped, you'd played a full round of telephone, and what came out the other end was a reasonable-sounding distortion of what was asked for.

A huge amount of what got called product management was standing in that chain, translating at each handoff, and absorbing the loss. That's real work.

It is also, if we're honest, project management wearing a product title: tracking who owes what to whom, keeping the game of telephone from completely garbling the message, without necessarily having the technical or business judgment to catch distortion before it shipped.

That is the work AI is actually good at: drafting the spec from the business conversation, keeping intent consistent from requirement to ticket, and catching gaps between what was asked and what got built before it ships.

The coordination layer, the part of the job that was pure telephone relay, is the part that's genuinely at risk. Not because it was not valuable. Because it was never supposed to be permanent.

🧭 What doesn't get automatedLink to this section

Tal Raviv, an IC PM who's built his whole practice around staying hands-on instead of climbing into management, wrote the piece circulating as "Product manager is an unfair role. So work unfairly." His line has stuck with a lot of people: the PM is often just "the cushion for organizational dysfunction." Read it here.

Raviv's piece is not really about AI replacing anyone. It is about building systems so the job stops eating one person alive.

One of his sharpest points is turning the whole team into mini-PMs, so initiatives keep moving without a single translator in the room.

That is the same shift from another angle. If everyone on the team can hold a piece of the business-to-ship chain without waiting for relay, you need less telephone.

What does not automate is judgment: the ability to catch a bad translation before it becomes three weeks of wasted engineering.

⚙️ What actually changed underneath thisLink to this section

AI did not just remove a layer of coordination. It shortened the test cycle enough to embarrass anyone using length as an excuse.

You used to say, "give us two sprints for a spike." Now a working prototype can happen in an afternoon.

That collapses the cost of finding out if an idea is right, which rewrites build-vs-buy. The safe default is not automatically safer anymore, because building to learn got cheap.

Speed did not just go up. The shape of the decision changed. Anyone whose whole value was standing in the gap between people who could move fast and people who could not just lost ground to the thing that closed the gap.

🪓 The stack, and what you bring to itLink to this section

If you are anywhere in that business-to-spec-to-UX-to-development chain, the honest question is not "am I in the doomed half." It is: what do I personally bring to this stack that is not just relay?

The best product people, designers, and engineers have already been blurring these lines for years: PMs who can sketch and read a query plan, engineers who understand churn, designers who sit in on sales calls.

AI did not start this convergence. It expanded who it applies to and compressed the timeline.

This role is not a shallow generalist. It is closer to a multi-level expert: deep in one discipline, with working fluency in the other two, enough to make a call and defend it.

Either the engineer moves up into market and outcome thinking, or the designer moves out into technical and business constraints. Both directions land in the same place.

That place is someone who can hold user experience, technical feasibility, and business outcome together in real time instead of reconciling three separate meetings a week later.

🧱 What this means if you are formalizing a product functionLink to this section

This changes what you hire for: not a requirements-writer with a product title chasing tickets through a chain, but someone who can stand beside engineering, understand what is possible this week, and know whether to build, buy, or skip.

That person is harder to find than a good coordinator. It is also the only version of the role worth building around three years from now.

Coordination is going away regardless of title. What replaces it is whoever can stand at more than one end of the chain at the same time.

#ProductManagement #AIStrategy #AIAdoption #ProductStrategy #Leadership #FutureOfWork #AIFirst