Competencies AI Commoditizes vs Elevates
Let me make this concrete by mapping it onto the actual twelve competencies of the Atlas. Not "AI changes everything," but which parts, and in which direction. Reforge framed the strategic move as moving to higher ground: as the mechanical work floods, you climb to the work that does not. This note is the map of where higher ground is.
#The map
AI commoditizes the production of artifacts and the retrieval of information. It elevates judgment, taste, framing, and influence, the things that decide whether the artifact was worth producing.
| Competency | What's getting cheaper | What's getting more valuable |
|---|---|---|
| Feature Specification | Drafting the PRD; formatting; first-pass user stories | Choosing the outcome; deciding what not to build |
| Product Delivery | Status reports, ticket grooming, coordination overhead | Trade-off calls; explaining the "why"; unblocking |
| Product Quality | Surfacing bugs and triage suggestions | Defining what "good" means, now via evals |
| Fluency with Data | Querying, charting, first-pass analysis | Causal reasoning; asking the right question; acting |
| Voice of the Customer | Synthesizing feedback across thousands of inputs | Knowing which signal is real and what to do about it |
| User Experience Design | Generating UI variations and flows in seconds | Taste; coherence; designing for failure modes |
| Strategic Impact | First-pass market and competitive scans | The actual strategic bet and its sequencing |
| Product Vision and Roadmapping | Roadmap drafts, formatting, options generation | Vision; saying no; the burden of choice |
| Stakeholder Management | Drafting updates, meeting notes, recaps | Trust, persuasion, conflict resolution |
The left column is converging toward zero marginal cost. The right column is appreciating.
#Why the premium on judgment is rising
Reforge put it plainly: the premium on judgment and taste has never been higher. The reason is the counterintuitive twist, which is that AI does not reduce complexity, it multiplies it. When you can generate ten roadmaps, twenty prototypes, and fifty spec variants in an afternoon, the scarce skill is not generation. It is choosing. The burden of choice lands on the PM, and choosing well is Product Sense and Judgment.
Lenny's metaphor lands here: PMs become editors of super-intelligent suggestions. The model produces; you decide what is good, what ships, and what gets killed. Editing takes more taste than writing, not less. Anyone who has edited knows it.
#The durable core
Strip away everything AI touches and what is left? Lenny calls it the people and glue work: influence, taste, vision, the judgment to frame the right problem. Marty Cagan's point reinforces it. The four big product risks (value, usability, feasibility, viability) predate AI and outlast it. AI helps you test risk faster. It does not decide which risk matters.
"Judgment is elevated" does not mean "ignore the tools." The PM who refuses to touch AI because their value is "strategy" is the one who gets out-executed by a PM with equal judgment and ten times the throughput. You climb to higher ground and you use the tools. Both.
#Continue Reading
- How AI Is Reshaping Product Management for the lifecycle view of the same shift.
- The PM Competency Model for the full twelve this maps onto.
- The 70% Problem for why the last 30% is where judgment concentrates.