What will product management look like in 2027?
The core list of assumptions and questions to think about as we explore product management in 2027
We spent an hour this month with a few hundred product people during our free monthly webinar on the future of product management. Not where product management lands in five years but rather what the job looks like six to twelve months from now.
Warren Buffett's line fits the moment: only when the tide goes out do you discover who has been swimming naked. AI pulled the tide out on software development. Build capacity was the constraint nearly every product process was designed around and focused on. It isn't anymore, and everything else that was slowing us down is now visible.
The shift, in one sentence
Product management in 2027 will be less about making things and more about creating the conditions for better decisions:
clearer problem framing
faster learning cycles, which is what Lean and Agile were always chasing
richer feedback from actual humans
alignment that holds across roles whose edges are blurring
None of these are new skills. They have moved from being perceived as overhead around the job to being the job itself.
1. The unsolved problem is the team workflow
Most of us have built a personal AI workflow by now like a chief of staff agent, a prototyping setup, a second brain. That works well in a company of ten.
Almost nobody has built a team workflow. You cannot hand everyone their own agents and hope they play nicely together especially at scale in the enterprise.
Artifact creation is cheap. When an engineer walks in with a clickable mockup, telling them to stop is the wrong answer.
Role definitions are loosening. Defending them is the fight the design community picked fifteen years ago when product managers started making wireframes, and it went badly then too.
These jobs were never the artifacts anyway. A designer understands the user's mental model. A product manager synthesizes the input and makes the call.
We asked the room what product managers should stop treating as proprietary. Their answers were surprising and, in our opinion, on point. Things like defining KPIs, requirements gathering, feature definition, owning product context are now shared contexts the whole team needs to do good work (hint: they’ve always needed these things).
2. Not all friction is waste
AI Transformation programs go looking for waste. How do we use AI to reduce friction, inefficiency, meetings that feel unproductive. Some of that friction really is waste. Some of it is how a group makes a decision.
We ran an OKR program where a manager kept trying to cut short the messy alignment conversations in the workshops. It turned out, those conversations were the work. The clarity and agreement the team needed happened inside them.
Facilitation becomes core product leadership work. You cannot hand it off to whoever happens to be running the workshop.
Decisions have to become explicit. When anyone can generate the artifact, somebody has to say which decisions matter, who makes them, and who is qualified to make them.
3. We ship daily. Do we listen daily?
Build cadence has collapsed to days, sometimes hours. Feedback cadence has not moved at all. The sensing half of Sense & Respond is badly out of step with the responding half.
The tempting shortcut is synthetic feedback. Generate AI personas, test against them, call it validation. We are not buying it. Anything AI hands you is a pile of Assumptions, and turning those into Hypotheses you test with real people is still the job of a product manager in 2027.
The better opportunity is feedback tooling built for places the mass-market research tools never reached. We used an example about one team that we know ships software onto offshore drilling platforms. To visit one you take a helicopter, and to board the helicopter you need scuba certification. That problem is finally solvable.
4. Roles dissolve, jobs migrate
Lou Rosenfeld told us UX is now the second most popular minor among business students at University of Michigan, behind entrepreneurship. Students treat it as a skill set they carry into whatever they end up doing. "Everyone becomes an AI builder" works the same way. It’s a mentality every role picks up. Nobody's business card is going to say builder.
So when a product manager moves upstream into framing, is that destruction? It is destruction of a role, not of a job. Goldman Sachs used to run night typing pools in the 80’s. Those jobs are gone. The work migrated elsewhere in the value chain, and this will too.
Two things from the Q&A worth keeping
Outcomes when agents talk to agents. Lagging indicators are still human behaviors. Agent activity often predicts value. Just don't mistake it for the value.
The optional skill that is now mandatory. Storytelling. When everyone arrives holding their own AI-generated output, the work that moves forward belongs to whoever can explain why it matters.
Take these to your next team meeting
What are we still treating as proprietary to one role?
Where is the mess in our process doing work we would lose by automating it?
How do we get human feedback at anything close to the speed we ship?
Which parts of our job titles are turning into skills anyone on the team can bring?
Watch the full session
The whole conversation, live Q&A included, is up now: Product Management in 2027