638 PMs on AI transformation — product management survey

In December 2024, researcher David Haberlah published a prediction about how AI would change product management tasks, mapping expected automation against a formal software product management reference framework. Fifteen months later, he returned to test the prediction against evidence.

The dataset

The source material is unusual for this kind of analysis: 349 newsletters and 289 podcast transcripts from Lenny Rachitsky’s newsletter and podcast, spanning June 2019 to March 2026. AI-related content across that archive rose from 4% of output in early 2023 to 67% by Q1 2026. The shift in what practitioners actually talk about is itself a finding.

What changed

The 2024 prediction held in its broad shape. Operational PM work — generating first drafts of PRDs, writing tickets, synthesizing interview notes — is automating at pace. Tasks that require judgment, organizational trust, and an ability to read context have held or gained in relative importance.

Specific cases give texture to the pattern. Duolingo moved from producing 100 courses over 12 years to 150 courses in roughly one year by applying AI across content production. Dan Shipper’s company Every runs five products at seven-figure revenue with a 15-person team where, in Shipper’s words, “no one is manually coding anymore.”

Non-determinism has become a first-order design constraint. PMs building AI-powered features now need evaluation frameworks and golden datasets, not just acceptance criteria. Aman Khan of Arize AI puts it directly: prompts are code artifacts that can regress, and teams that treat them otherwise will discover failure modes at the worst moment.

What endures

Across 638 practitioner voices, three qualities came through as more valuable in an AI-augmented environment, not less: conviction about what to build, taste in evaluating outputs, and the ability to influence people without formal authority. The article introduces a “PM Compass” framework mapping these enduring skills against the new competencies now expected of product managers.

Who it is useful for

Senior PMs and product leaders thinking through how the PM role is evolving, and anyone making the internal case for what product management should look like on an AI-first team. The article is long — about 16 minutes — and includes a full bibliography of primary sources drawn from Lenny’s archive.