The PM Role Is Splitting. Most CPOs Haven't Noticed Yet.

Execution-layer work is already gone. You just haven't restructured around it yet.

There is a clear line between two types of PM work. On one side: anything that takes structured inputs and produces structured outputs. User interview summaries. PRDs. Sprint scope documents. Release notes. Prioritization scoring. This is execution-layer work.

Every one of those tasks is being automated by the top 10 percent of PM teams right now. Not piloted. Not experimented with. Done. Regularly. At a fraction of the time cost.

The data reflects this. 85 percent of product leaders say they are investing in AI tools. Only 2 percent are investing in PM talent development to match. That is not a pipeline problem. That is a structural mismatch. The tools have moved faster than the role definition, and most teams are running an outdated job description with a modern toolset bolted on top.

What that looks like in practice: a senior PM spending four hours writing a PRD that an AI could draft in twelve minutes, then spending another two hours in a meeting that exists because nobody has defined what decisions actually need a human in the room.

Strategy-layer work is getting more valuable, not less.

Here is the uncomfortable part of this conversation: the answer is not to celebrate AI freeing up PM time for "more meaningful work." That framing lets everyone feel good without changing anything.

The real question is what that freed-up time is actually worth. And the answer depends entirely on whether your PMs can operate at the strategy layer.

Framing the right problem. Making trade-offs under genuine uncertainty. Building organizational alignment across a leadership team with competing priorities. Deciding what not to build, and defending that decision for six months. These things cannot be delegated to an agent. Not because AI isn't capable enough yet, but because they require context, relationships, and judgment that live inside people, not models.

Melissa Perri put it directly in Escaping the Build Trap: "The product manager's most important job is to work with the team to create the right product — not to manage the process of building what stakeholders ask for." That was true before AI. It is more true now, because the cost of defaulting to the wrong job has gone up significantly.

The market is pricing this in. AI-fluent PMs who operate at the strategy layer are commanding a 28 percent compensation premium over peers doing the same role title. That number tells you everything about where value is concentrating.

Restructuring requires a new performance framework, not a new tech stack.

This is where most CPOs stop. They roll out AI tools, run a workshop, and call it a strategy. The role boundaries don't change. The performance criteria don't change. PMs are still measured on documents produced, tickets closed, and velocity delivered.

That framework rewards execution-layer output. It actively works against the shift you say you want to make.

If you want your PMs operating at the strategy layer, you need to evaluate them on strategy-layer outputs: decisions made and the quality of the reasoning behind them, problems framed sharply enough that the team built the right thing, trade-offs surfaced early enough to matter. Not PRDs written. Not story points shipped.

The teams getting this right are not running AI pilots. They are redesigning role boundaries. They are treating AI leverage as a first-class resource for PMs, the same way they treat engineering capacity. And they are having explicit conversations about what a PM is now accountable for — which means also being explicit about what the PM is no longer accountable for.

One thing to do this quarter

Pick one execution-layer task your PMs do every week. Scope it, hand it to AI, and measure the output quality honestly. Then take the time you recover and define what a strategy-layer deliverable from that PM actually looks like. Write the evaluation criteria for it.

That is not an AI pilot. That is the beginning of a structural decision. The rest follows from there.