The Skills That Got More Valuable — and the Ones AI Just Made Cheap

The Skills That Got More Valuable — and the Ones AI Just Made Cheap. Here's the uncomfortable truth about the product management craft in 2026: a chunk of what used to make you look competent is now a free feature in ChatGPT. The first draft PRD, the competitive teardown, the tidy meeting summary — all of it can be produced in seconds, at roughly the quality of an average PM having an average day. That sounds like bad news. It's actually a re pricing. The skills didn't disappear; they split into two piles. One pile got cheap. The other got more valuable than ever. Knowing which is which is the whole game now. The market is voting with job postings Start with the loudest signal. McKinsey found that demand for AI fluency — the ability to actually use and direct AI tools — grew roughly sevenfold in two years, faster than any other skill in US job postings. The number of workers in occupations that explicitly require it jumped from about 1 million in 2023 to 7 million in 2025, and three quarters of that demand sits in computer/math, management, and business roles. Product is squarely in the blast radius. [1] This isn't "learn to prompt." It's a baseline expectation, the way "comfortable with spreadsheets" became invisible a decade ago. The PMs who treat AI as a novelty are quietly being sorted below the ones who treat it as a power tool. What PMs say got more important When Productboard surveyed 379 product professionals at 500+ person companies, it asked which skills are rising in value as AI handles the tactical work. People picked an average of 3.5 each, and the top four are revealing: none of them are things AI does well. [2] Notice the pattern. Data literacy, synthesis, systems thinking, strategy — these are the parts of the job that require taste, context, and the ability to hold a messy whole in your head. AI can draft the artifact, but it can't decide whether the artifact is pointed at the right problem. The "produce vs. think" split The Lenny's Newsletter productivity survey (1,750 tech workers) makes the divide almost mathematical. PMs have enthusiastically adopted AI for producing — writing PRDs (21.5%), making prototypes (19.8%), drafting communications (18.5%). But for thinking — the upstream work of figuring out what to build — adoption craters: user research sits at just 4.7%, generating roadmap ideas at a rounding error 1.1%. [3] Read that as a map of where value is moving. The bottom of the chart — research, prioritization, deciding — is exactly where humans still hold the pen. That's not because AI can't help there; it's because those problems don't have clean, checkable outputs. A "pretty good" PRD is useful. A "pretty good" strategy that's quietly wrong is expensive. Judgment is the new alpha The cleanest way I've heard it put comes from Adam Judelson, former Head of Product at Palantir: My Take Your Summary So the craft in 2026 has two new entry fees. One: genuine AI fluency, so you're not paying full price in time for work that should be near free. Two: the judgment to know which problems are yours and which belong to the machine. AI raised the floor on output and, in doing so, raised the bar on what counts as good. The good news hiding in all of this: the parts of the job most people actually got into product for — the strategy, the customer obsession, the calls no one else will make — are the parts that just appreciated. The paperwork is what got automated. Next in this series: the job market itself. Openings are at a three year high — so why does it still feel brutal out there? Previous Parts: the builder (Part 1)