AI and the Modern PM
Evals, vibe-coding, AI product sense, and the competencies AI is rewriting.
This is the domain I care about most, because it is the one actively rewriting the job description while you read this. Everything else in the Atlas still holds: the twelve competencies, the ladder, the frameworks. AI does not replace that spine. It changes which parts of it are scarce. What follows is my read, as someone currently building products with AI, on where the job is going and what to do about it.
#The thesis
The PMs who win the next five years are the ones who become builders again. The mechanical core of the job (drafting specs, summarizing research, writing status updates, basic analysis) is commoditizing fast. The premium is moving to judgment, taste, problem framing, and the new technical literacy of shipping probabilistic systems. AI does not reduce the work. It raises the floor and the ceiling at the same time.
Lenny Rachitsky argues PMs are best positioned to thrive in the AI era, and I think he is right, conditionally. The condition is adaptation. The PM whose value was "I write the doc and run the meeting" is in trouble. The PM whose value was "I decide what is worth building and why" just got a force multiplier. Same title, opposite trajectories.
I will be blunt about the hype. "PM is dead" is clickbait (more in Myths and Misconceptions). The honest version is that the commodity PM is fading and the judgment PM is compounding.
#What's actually new
A few skills crossed from nice-to-have to table stakes. The biggest is evals. OpenAI's Kevin Weil calls writing evals a core PM skill, and I agree: in my view they are the new PRD. Said carefully, a PRD (or an AI PRD) still describes the intent, while the evals prove the behavior. They complement each other rather than cancel each other out. If you take one thing from this domain, make it Writing Evals for AI Products.
#The map
| Start here | Then |
|---|---|
| How AI Is Reshaping Product Management, the lifecycle shift | Competencies AI Commoditizes vs Elevates, mapped onto the twelve |
| The PM's AI Literacy, the "just enough" technical bar | AI in Discovery and Research, synthesis at speed |
| AI in Prototyping and Delivery, vibe coding, builders again | Writing Evals for AI Products, the new core skill |
For the specialist track and how to build well:
- The AI Product Manager for the three-way taxonomy and what an AI PM actually owns.
- Building AI Products for designing on top of probabilistic systems.
- Designing for Trust and Probabilistic Systems for hallucination, guardrails, and human-in-the-loop.
- The Unit Economics of AI for why inference cost is not SaaS margin.
- The 70% Problem for where AI accelerates and where judgment concentrates.
- AI Ethics and Governance for the PM as guardian of trust.
#A note on the numbers
Throughout this domain I cite salary and adoption figures: AI PM comp, the "20% of PM jobs" claim, and so on. Treat all of it as directional. The market is moving monthly, sources disagree, and Levels.fyi is a snapshot rather than gospel. I use the numbers to show direction and magnitude, not to promise a specific paycheck.
#Continue Reading
- How AI Is Reshaping Product Management for the single best place to start.
- The PM Competency Model for the spine AI is reshaping, not replacing.
- Product Sense and Judgment for the durable core that gets more valuable.
- Fluency with Data for the adjacent competency AI literacy extends.
- Myths and Misconceptions for where I separate the AI hype from the real shift.