Skip to content

Start here

Your first AI week as a PM

Most "AI for PMs" advice hands you a list of 200 tools and wishes you luck. This is the opposite: five days, one small win per day, each one a thing you can actually finish between meetings. Total time is about 2.5 hours across the week — and by Friday you'll have replaced your two worst time sinks.

Nothing here needs budget approval: every step works on a free tier or a $20/month plan you may already have.

  1. Monday

    ~15 min

    Stop taking notes in meetings

    Capture is the one thing worth automating first, because it's passive — set it up once and every later step gets free input. Nothing else this week works without it.

    1. 1. Pick your capture tool: Granola if you want notes with no bot joining the call, Otter.ai if you want a generous free tier and speaker labels.
    2. 2. Connect it to your calendar so it joins (or listens to) meetings automatically.
    3. 3. Do nothing else. Go to your next meeting and actually pay attention — that's the win.

    First click: Open the Granola or Otter page and start the free tier.

  2. Tuesday

    ~20 min

    Let a skeptic tear apart your next spec

    The highest-leverage AI use for a PM is not writing the doc — it's attacking the doc before your staff engineer does. One prompt, and you walk into review already knowing the three holes.

    1. 1. Take a PRD or spec you consider finished (not a rough draft — the point is stress-testing something you'd otherwise ship).
    2. 2. Copy the Red-team your PRD prompt and run it in Claude or ChatGPT.
    3. 3. Fix only what you agree with. The prompt is forbidden from rewriting your doc, so every critique quotes the exact line it targets — you stay the author.

    First click: Hit Copy on the red-team prompt.

  3. Wednesday

    ~45 min

    Turn old customer calls into evidence you can cite

    This is the day the week pays for itself. Interviews you already ran, sitting unwatched, become a ranked list of themes with verbatim quotes you can put in front of an exec.

    1. 1. Grab transcripts from your 3 most recent customer calls (Monday's tool has been collecting them; older recordings work too).
    2. 2. Follow the interview-backlog playbook end to end — it chains capture → synthesis → an evidence-backed PRD section.
    3. 3. Do the spot-check step. Verify three quotes against the source. That 10 minutes is what makes the output defensible instead of merely confident.

    First click: Open the playbook and read the chain before you start.

  4. Thursday

    ~30 min

    Answer one data question without joining the queue

    You don't need to become an analyst. You need to stop waiting two weeks for a number that decides something today.

    1. 1. Pick one real question you're currently blocked on ("where do trial users drop off?").
    2. 2. Try it in your analytics tool's AI first (Amplitude, Mixpanel), or take a CSV export to Julius with the CSV interrogation prompt.
    3. 3. Turn the answer into a one-page decision memo — it forces you to write what the data does not say, which is where most analysis goes wrong.

    First click: Write down the one question, then open the CSV prompt.

  5. Friday

    ~40 min

    Ship your weekly update in 40 minutes

    Monday's capture has quietly been collecting your whole week. Now it becomes the update you usually spend an evening on.

    1. 1. Pull your week of meeting notes (or run the weekly meeting digest first to extract decisions and owned actions).
    2. 2. Run the stakeholder update memo prompt — fixed format: Shipped / Slipped / Decisions needed / Risks.
    3. 3. Personally verify every "Decision needed" line. That's the ask you're making of a busy person; the prompt quarantines anything it couldn't confirm so you can't bluff by accident.

    First click: Copy the stakeholder update memo prompt.

Then what?

Pick the thing that's still eating your week. The playbooks each start from a specific struggle and chain the tools to kill it; the prompt library is there when you just want to steal something that works. And if you want to know why a tool made our list — or why we cut one — every verdict names a real weakness.

After that, staying current is one email a week: The Cut tests what's new and tells you what failed, so you don't have to run the sweep yourself. Weeks where nothing is worth your time, it says exactly that.