The PM Role Is Splitting in Two
AI isn't replacing PMs. It's splitting the role into two very different jobs.
TL;DR · The comfortable middle of product management is compressing. AI is commoditizing the administrative layer (briefs, summaries, ticket grooming, status updates) while judgment, taste, system thinking and product instinct stay expensive. The role is forking into two paths: Coordinator PM (manages the machine) and Force Multiplier PM (reduces complexity). Both are valid, but they're not the same job anymore.
There used to be a comfortable middle in product management.
Write clean PRDs. Run a tight sprint process. Keep stakeholders aligned. Synthesize research into slides that made people nod. Move tickets across boards. Be the connective tissue.
You could build a whole career in that middle. A lot of people did. It was safe, respected and reasonably well-compensated. The PM as professional translator. Taking messy inputs from every direction and producing organized outputs that kept the machine running.
That middle is compressing.
Not because the work stopped mattering. Because a lot of it stopped being hard to do.
What's getting cheaper
AI can write a first draft of a brief in seconds. Summarize a research call. Cluster a thousand support tickets into themes. Clean up a backlog. Structure a decision log. Draft release notes nobody reads anyway.
Not perfectly. But well enough. The gap between "PM wrote this" and "AI wrote this while the PM got coffee" is shrinking fast.
The work that used to justify a PM's entire Tuesday now takes fifteen minutes and a prompt.
That doesn't make it worthless. It makes it table stakes. Like being good at Excel in 2010. Necessary. Not differentiating.
If your main value is producing documents and keeping a process running, you're competing with tools that don't get tired, don't have opinions about meeting frequency and never need a mental health day. That's not a fight you win by typing faster.
What stays expensive
Judgment. Taste. The ability to look at ten possible directions and know (not guess, know) which three are worth exploring and which seven are noise dressed up as opportunity. The willingness to commit to a direction and hold it long enough to learn something.
System thinking. Understanding how the product, the market, the user and the business fit together in ways that never show up in a Jira filter. Or a status update. Or that dashboard someone spent three weeks building that nobody looks at.
Product instinct. The thing that tells you a feature is going to flop before you ship it. Or that a weird data pattern in Tuesday's pull is the most important signal anyone will see all week. You can't prompt-engineer this. It comes from reps.
AI can generate options. It can't tell you which one matters.
These skills don't compress. They compound. And they're getting more valuable every month. Because when execution gets cheap, bad decisions ship faster too. The person who can say "not that, this" before the team burns a sprint on the wrong thing just became the most expensive player on the roster.
It's like giving everyone on the team a legendary weapon. Great. But if the raid leader doesn't know the fight mechanics, you're just wiping faster with better gear.
The two paths
The role is forking. Not hypothetically. Not in some future-of-work blog post. Now.
Coordinator PM. Manages the machine. Keeps alignment. Runs the rituals. Handles stakeholders. Tracks delivery. Still valuable in big orgs where coordination overhead is real and the alternative is twelve people building three different things. But the ceiling drops every quarter. If the most important thing you do is keep things organized, you're one reorg away from being a project manager with a different title and the same Confluence page.
Force Multiplier PM. Uses judgment, data, AI and direct product contact to collapse the distance between question and answer. Doesn't just manage the work. Shapes the bets. Pressure tests the logic. Finds signal before consensus catches up. Gets close enough to the product and the user to have opinions that are earned, not inherited from a dashboard someone else built.
One version manages complexity. The other reduces it.
Both exist. Both have a place. But they're not the same job anymore. And pretending they are is how you end up with a PM team where everyone's busy and nobody can explain what the product learned last month.
Why this matters now
Companies are going to start hiring for one of these while describing the other.
The job post will say "strategic product leader." The actual job will be scheduling standups and negotiating whether Wednesday's sync can move to Thursday. Or the post will say "execution-focused PM" and what they really need is someone who can figure out what's worth building before the team builds the wrong thing. Again.
If you've ever interviewed for a "strategy" role and the first question was about how well you craft a user story, you already know.
Meanwhile, the PMs who actually drive outcomes (look at messy data, see a bet, prototype something before anyone realizes it's worth testing) are getting harder to find. Not because they don't exist. Because the role was never designed to select for them. It selected for process fluency. Communication skills. Stakeholder management.
For a long time that was enough.
It's the RPG class problem. The game used to reward generalists who could do a little of everything (looking at you druids). Now the meta shifted and the hybrid builds don't scale. You're either speccing into impact or you're speccing into coordination. Both are valid classes. But if you think you're building one and you're actually building the other, you're going to wonder why nothing compounds.
What to do about it
Not a career guide. I'm barely qualified to run my own. But a few things I keep seeing matter.
Get closer to signal. Stop waiting for the research summary. Talk to users yourself. Read the raw feedback. Pull your own data. Use the product. The PM who's closest to reality makes the best calls. Every time. No exceptions. I have yet to see this not be true.
Use AI at the workflow level. Not as a chatbot you open when you're stuck. As infrastructure. Let it handle the synthesis, the formatting, the first drafts so you can spend your actual hours on the judgment calls that need a human with context and taste. The stuff AI is bad at is exactly the stuff that matters most. That's not a coincidence.
Build taste. You can't A/B test your way to product sense. It comes from using products obsessively, studying what works and developing a feel for why things land or don't. This is the part that doesn't automate. Probably won't for a long time. Maybe ever.
Prototype more. Not to impress anyone. To pressure test your own thinking. Specs describe what you hope will happen. Prototypes show you what actually does. That's a very different conversation to walk into.
Stop hiding behind process. Process is a tool, not a job description. If the most visible thing you did this week was facilitate a meeting, that's not a flex. That's a warning sign.
The question isn't whether AI changes the PM role. It's which version you're building toward.
The point
The middle was comfortable. Predictable. A lot of PMs built real careers there and there's nothing wrong with that. But the middle is compressing. The floor is rising. The work that used to be enough to justify the seat is getting absorbed by tools that do it faster and don't complain about the open office plan.
And the ceiling for the other version (the one closer to product, closer to signal, closer to outcome) is going up. Way up.
One side of the fork is becoming a support function. The other is becoming the highest-leverage role on the team.
Choose accordingly.