The Four Levels of AI Fluency

Everyone on the team uses AI. Almost nobody can show what changed because of it.

Short answer

The four levels of AI fluency for PMs are Unacceptable, Capable, Adoptive and Transformative. The deciding variable is whether someone used AI as a faster version of the old workflow or redesigned the workflow. Place people by asking what their team can do today that it could not six months ago, specifically because of how they use AI.

Key takeaways

  • Unacceptable means the work looks the same before and after AI: a faster keyboard.
  • Capable means repeatable systems, such as a prompt library, that you could teach another PM.
  • Adoptive means building infrastructure that other people rely on.
  • Transformative means the org works differently, and the thinking survives swapping every tool.
  • Use the interview question to place candidates, or audit your own answer.

Common questions

What are the four levels of AI fluency?

Unacceptable, where work looks the same as before; Capable, with repeatable systems you can teach; Adoptive, where you build infrastructure others rely on; and Transformative, where the org works differently because of you. Each level has specific tells.

How do I assess a PM's AI fluency in an interview?

Ask: what can your team do today that it could not six months ago, specifically because of how you use AI? A time saving claim is Level 1, a teachable workflow is Level 2, and something a colleague depends on is Level 3.

Can a PM drift up the AI fluency levels as tools improve?

No. A Level 1 candidate does not drift up to Level 3 as tools improve. The gap is mindset, not access, and it rarely closes without deliberate pressure. Use the levels to write a job scorecard or audit yourself.

CheatsheetLayout, Section, NumberedItem, Callout, } from "@/components/CheatsheetLayout";

The variable that actually separates people

The decisive variable is not intelligence or effort. It is whether someone treated AI as a faster version of the old workflow, or as a reason to redesign the workflow. That single choice explains most of the distance between two PMs with identical tool access.

The four levels

Unacceptable: the work looks the same before and after

Tells: Pasting a PRD in and asking "does this look good", then accepting whatever comes back. Using AI to summarise research and missing the insight. One tool for every job, with no view on which model fits which task. AI as a faster keyboard.

Capable: repeatable systems they can teach another PM

Tells: A refined prompt library for recurring work. Feeding 200 NPS responses in to surface the power user versus new user tension. Generating and pressure testing five solution directions in a week. A system, not a habit.

Adoptive: they build infrastructure other people rely on

Tells: Support tickets tagged by feature and churn risk, feeding prioritisation. Running their own SQL and shipping a working dashboard before engineering touches it. An automated digest that catches metric anomalies before the weekly review. The key signal is that someone else uses what they built.

Transformative: the org works differently because of them

Swap every tool and the thinking survives. Tells: PMs shipping low risk, high frequency changes to production, with the tradeoffs articulated. Quarterly discovery replaced by a continuous opportunity map, rather than sped up. No longer writing specs, instead setting the judgement framework the AI drafts against.

The one interview question

"What can your team do today that it could not six months ago, specifically because of how you use AI?" The quality of that answer places people faster than almost anything else. A time saving claim is Level 1. A teachable workflow is Level 2. Something a colleague depends on is Level 3.

Why it matters

A Level 1 candidate does not drift up to Level 3 as tools improve. The gap is mindset, not access, and it rarely closes without deliberate pressure. Use it when you are writing an AI PM job scorecard, or auditing your own answer to the Level 3 question. Once you have named the level, name the rung: The Three Ladders of AI Leverage turns the level into a concrete next move. Source: Lewis Lin, "4 Levels of AI Fluency in Product" (Apr 2026), adapting Zapier's four level framework. );