The PM Role Is Splitting in Two

What exactly has been automated

The operator layer of product management is gone, or nearly gone. I mean the work that used to fill a PM's week: synthesising user interviews into themes, writing tickets, generating first drafts of PRDs, pulling data into weekly status updates, maintaining prioritisation spreadsheets, drafting stakeholder communication.

These tasks required time, diligence, and a certain kind of pattern recognition. They were real skills. I am not dismissing the people who built their reputation doing them well. But those tasks are now AI-executable, and doing them manually is not a differentiator anymore. It is just slow.

What remains on the operator side is QA of AI output, which is a different skill entirely, and not one that justifies the same headcount model.

What the exposed PM looks like right now

There is a PM profile common in European SaaS companies, typically someone who rose through execution excellence. They are organised, rigorous, good at keeping five workstreams moving. Stakeholders trusted them because they always had the answer, always had the doc, always had the update ready.

In 2024 that was a strong PM. In 2026 it is a risk profile.

Not because they are bad at their job. Because the tasks they optimised for have been absorbed into the tools layer, and the skills they deprioritised — navigating ambiguity, building cross-functional alignment without formal authority, translating a business strategy into a product direction that the team actually internalises — are now the differentiators.

Melissa Perri puts it clearly in *Escaping the Build Trap*: "The best PMs I know are not the ones who can manage a backlog. They are the ones who can make a strong argument for what not to build." That sentence lands differently now than it did five years ago.

The restructuring lens you actually need

When I work with CPOs rethinking their team structure, I push them to ask a different set of questions in reviews and interviews.

Not "How do you prioritise your backlog?" but "Tell me about a time you had to change what the business thought it wanted." Not "How do you run discovery?" but "How do you make a call when the data is inconclusive and the stakeholders disagree?"

The strategy-first PM you are looking for can do several things that are genuinely hard to automate. They can hold a customer conversation and come back with a point of view, not just a summary. They can walk into a room where engineering, commercial, and executive priorities are in conflict and leave with something close to alignment. They can translate a company-level objective into a product direction without needing someone above them to pre-chew it.

Assess for judgment and cross-functional influence directly. Give candidates a real ambiguous scenario from your business. See how they think through it, not how clean their framework is.

On structure: my experience is that the right model in an AI-augmented product team is fewer PMs owning broader surface area, with AI handling the operator layer explicitly. If your current ratio is one PM per two or three engineers, you should be questioning that assumption. The span of ownership can increase when execution overhead decreases. What you need is not more PMs. You need the right kind.

The part that applies to you too

Here is the uncomfortable part.

The same split happening in the PM role is happening in the CPO role. The CPOs who define their own value through delivery oversight — who are in the room because they know the roadmap cold, who are trusted because they always have the update ready — are in the same position as the execution-first PM they are about to restructure around.

Strategic translation, board-level narrative, organisation design, building the conditions where strong product thinking can happen at scale: that is the irreplaceable work of a CPO in 2026. If your calendar is mostly status reviews and escalation calls, that is worth looking at honestly.

The restructuring conversation is not just about your team. Start there, but don't stop there.