How to use AI as a product manager

Most advice on this question is a list of tools. The better starting point is a list of your own tasks, because the gains from AI depend far more on which work you point it at than on which product you pick. In Generative AI at Work, an NBER study of 5,179 customer support agents, Erik Brynjolfsson, Danielle Li and Lindsey Raymond found an AI assistant raised issues resolved per hour by 14% on average, "including a 34% improvement for novice and low-skilled workers but with minimal impact on experienced and highly skilled workers." Same tool, very different payoff depending on the person and the task.

The second reason to measure before you scale is that people misjudge AI's effect on their own work. In a 2025 randomised trial, METR found experienced open-source developers took 19% longer on real issues when allowed AI tools, and noted that "developers expected AI to speed them up by 24%, and even after experiencing the slowdown, they still believed AI had sped them up by 20%." METR calls that result a snapshot of early-2025 tools in one setting and has since published newer data, so it is not a verdict on AI. It is a warning that your sense of being faster is not evidence.

How do you audit your PM workflow for AI?

Block one hour. Open your calendar, your sent mail and your docs from the last two weeks, and write down every task you did more than once. For each, note roughly how many hours a week it takes and whether it ends in a decision or just moves information from one place to another (collecting updates, reformatting notes, summarising calls).

Keep the five that take the most time. Then score each on one extra question: does AI do this kind of task well? The jagged frontier explains why that question matters. AI is strong on some tasks that look hard and weak on some that look easy, so you judge it task by task.

Here is an illustrative audit for a PM on a B2B product. The hours are an example, not data:

Recurring task	Hours a week	Decision or information-moving	AI fit	Priority
Weekly status update to leadership	2	Information-moving	High	1
Synthesising customer call notes	3	Information-moving, feeds decisions	High	2
Backlog grooming and ticket writing	3	Mixed	Medium	3
Preparing for roadmap review	2	Decision	Low for the call, high for prep	4
Resolving cross-team priority conflicts	2	Decision	Low	Keep manual

The status update wins because it is frequent, fully information-moving and easy to check. The priority conflicts stay manual because the value is in your judgment and your relationships. The roadmap review splits: AI gathers and structures, you decide.

Why should you build your AI stack like an MVP?

Because each layer of AI tooling adds cost, setup time and something new that can break, and you only learn whether a layer is worth it by using it. Treat your AI setup the way you would treat a product: ship the smallest version, measure, and add the next piece only when the current one has shown value.

Six steps, each one earned by the step before:

1. Break the task down. Take your priority 1 task and write its steps. For a status update: collect updates, pull metrics, draft, adjust tone for the audience, send. AI will help with some steps, not all.
2. Use a general assistant. Run the AI-friendly steps in whatever assistant you already have. No setup, no new subscription.
3. Check the value. Two weeks, one line per run: minutes saved, errors caught, anything new that could go wrong. This is the step most people skip, and the METR result is why it matters.
4. Save the setup. Once the prompt stops changing, turn it into a reusable assistant with standing instructions and reference files. Anthropic's help centre describes how Claude projects let you upload documents to a project's knowledge base and "define project instructions for each project to further tailor Claude's responses," and Gemini Gems work on the same principle.
5. Add a specialised tool only for a clear gap. A meeting recorder, a research tool or an analytics assistant earns its place when the general assistant cannot reach the data or format you need. See best AI tools for product managers for what each category does.
6. Build custom last. Retrieval over your own documents or an agent that acts across systems comes only after steps 4 and 5 have hit a real limit. RAG for product managers explains when retrieval is the right fix.

Most PMs get most of the benefit from steps 2 to 4. Step 6 is where people start when they read a launch announcement, and where they stall.

What kinds of work does AI do best for a PM?

Three kinds of work cover almost every PM use of AI, and each needs a different check:

• Drafting: first versions of updates, PRD sections, release notes. Check for claims you did not supply.
• Transforming: rewriting a spec for sales, turning call notes into a summary, translating tone for executives. Check that nothing was dropped or invented in the conversion.
• Analysing: clustering feedback, comparing options, spotting gaps in a plan. Check the conclusions against the raw material, because this is where confident wrong answers hide.

For task-specific walkthroughs, see AI for stakeholder updates and AI for customer interviews.

What is the most common mistake PMs make with AI?

Automating before understanding. A PM builds a clever multi-step setup for a task they have never timed, cannot show it saved anything, and quietly goes back to doing it by hand. The audit and the value check are dull, and they are the part that makes the rest stick.

The strongest objection to this method is that it is slow: two weeks per step while AI tools change monthly. The answer is that the audit and the measurement transfer across tools. When a better model ships, you swap it into a step you already understand and compare it against numbers you already have, instead of starting over. If you want to take the setup further, build a personal AI operating system covers how to connect the pieces.

Where can you build this system with guidance?

10x Productivity with AI is a 1 week Builders Camp bootcamp with 2 live sessions, taught by Andre Albuquerque and part of the AI Agentic Builders Expert Track and the Product Delivery Specialist Track. Its published topics include AI workflow design (choosing the right tasks, inputs and tools so productivity gains are real and measurable), writing and editing at speed, meeting and notes automation, research and synthesis, task automation with guardrails, and safe usage habits. Its practical challenge asks you to audit a PM's recurring week in 48 hours and show concretely where AI helps and where it does not.

See the 10x Productivity with AI bootcamp