Why AI Agents Will Make Some Product Managers Irrelevant (And Others Indispensable)

Every few years, someone publishes a piece arguing that product managers are about to be replaced. By designers. By engineers. By data scientists. By founders who just do it themselves.

The current version of that argument is about AI. And this time, it's more specific than it used to be.

It's not that AI will replace product managers as a function. It's that AI is creating a K-shaped split in the PM job market, where two groups are pulling ahead while a large middle gets squeezed out. The PMs who are becoming irrelevant aren't the bad ones. They're the ones whose entire value lives in the parts of the job that AI now does faster.

What AI Has Already Replaced

Let's be direct about what's gone or going.

Spec writing as a core skill. A PM who spends 40% of their week writing PRDs, user stories, and acceptance criteria is a PM whose value is declining. AI writing tools produce working first drafts of all of these in minutes. The value was never in the document. It was in the thinking behind it. PMs who mistook one for the other are in trouble.

Synthesis from known sources. Competitor analysis, market research summaries, feature gap analysis from existing data, summarizing customer feedback across tickets: all of this is faster with AI. In my experience, a junior PM using AI can produce in two hours what used to take a full week of research work.

Meeting facilitation as a differentiator. Running standups, writing meeting notes, distributing action items. These were never high-value activities. They're now even lower value because AI handles most of them automatically.

Ticket management and backlog grooming. If your primary contribution to a product team is maintaining the Jira board and writing ticket descriptions, that work is being automated. Not eventually. Now.

Research from Agents Today describes this as the disappearance of the "traditional mid-level generalist PM role." The tasks that defined entry-level and mid-level PM work for the last decade are being compressed by AI, which means the ladder that used to exist for climbing from junior to senior PM is getting shorter.

What AI Cannot Do

Here is where the picture becomes more interesting.

Judgment under genuine ambiguity. AI is extraordinarily good at synthesizing information from known sources. It is bad at the moment when there are no good sources, when you're deciding something that hasn't been decided before, when the data contradicts the customer, when the founder's instinct is different from the analyst's model. The PMs who are becoming more valuable are the ones who can hold that kind of uncertainty and still make a call.

Earning trust across a dysfunctional organization. No AI agent is going to get the engineering lead and the design lead into the same room and get them to agree on scope. No AI is going to navigate the political dynamic between two competing stakeholders and come out with alignment. Cross-functional influence without authority is a human skill. It remains one of the highest-leverage PM skills precisely because it cannot be automated.

Defining the right problem. AI is exceptionally good at solving well-defined problems. It is not good at questioning whether the problem being solved is the right one. The PM who walks into a roadmap meeting and says "I think we're optimizing for the wrong metric" is doing something that requires context, relationships, customer knowledge, and the willingness to be unpopular. That's not going away.

Customer proximity. Understanding what a user actually means when they give feedback, as opposed to what they literally say, requires human judgment built from hours of direct interaction. The PM who has talked to 50 customers this quarter understands something that cannot be learned from a transcript summary.

The K-Shape in Practice

The data from hiring markets is specific: AI-focused PM specialists are earning approximately 35% more than their traditional counterparts. 75% of employers say they struggle to find qualified AI product managers. Two-thirds of business leaders say they would not hire PM candidates who lack AI skills. And 71% of hiring managers say they would prefer a less experienced candidate with strong AI skills over a more experienced candidate without them.

That last number is the one that matters most for anyone mid-career. The experience premium is being discounted when it comes with no AI fluency attached.

The two PM profiles gaining ground are distinct but both AI-literate. The first is the AI product specialist, someone who builds AI-powered products and understands LLMs, agents, and evaluation at a technical level. The second is the AI-augmented generalist, a PM who uses AI across their entire workflow and therefore moves at a speed and coverage that a non-AI-augmented PM simply cannot match.

The profile losing ground is the PM whose primary productivity comes from personal manual effort on tasks that AI completes in minutes.

What I Think This Actually Means

I've been thinking about this from the lens of what makes a PM genuinely useful to a team, not just present.

The PMs I've seen create the most value in my experience are the ones who do something that requires specific human judgment: they understand the customer in a way that isn't in any database, they build trust with engineering in a way that makes hard conversations easier, they read between the lines of what a user says to find the actual problem. None of those things are in danger of being automated.

What is in danger is the PM who has built a career on being the person who writes things down, runs the process, and synthesizes what others have already said. Those tasks are being redistributed.

The honest version of the question isn't "will AI replace product managers?" It's "do I currently spend most of my time on things AI can now do?" If yes, that's not a threat to the PM role in general. It's a specific signal about which direction to grow.

The PMs who will be indispensable in the next five years are the ones who treat AI as a way to free up time for the high-judgment work, not as a threat to be worried about. The job is getting better for the people who understand that distinction.