How I use Claude + Notion to run my week without losing my mind
I'm a Senior Principal PM. I manage a platform, work across three time zones, and sit in more meetings than I'd like. For a long time, staying on top of everything meant working longer. Then I started building systems instead.
This isn't about replacing thinking with AI. It's about removing the friction between "something happened" and "I know what to do next." These are the three AI workflows for product managers that changed how I work — here's the system I actually use, every week.
The problem I was solving
As a PM, I sit at the intersection of customer feedback, engineering capacity, and business priorities. The raw material is everywhere — support tickets, community forums, Slack threads, customer calls, competitive research. The hard part isn't finding signal. It's synthesizing it fast enough to act on it, and then turning that synthesis into specs, stories, and content that other people can actually use.
Three workflows changed how I work.
The principle behind all of them: AI doesn't replace PM judgment — it compresses the distance between insight and output. The thinking is still yours. The formatting, the first draft, the structure — that's where the time goes, and that's what AI handles.
The three workflows
Meeting notes → Notion, in my voice
AI meeting notes are now part of every PM's toolkit — here's how I've set it up to actually sound like me. After every Zoom call, I paste the AI summary into Claude with a simple prompt. It strips the noise and formats the output as a clean Notion entry — decisions, blockers, next steps — with @Name tagging so owners are clear. Takes about 90 seconds. I paste the result directly into the right project or client page in my workspace. No reformatting, no rewriting. It sounds like me because I've tuned the prompt to match how I actually write.
Scraping signal → backlog candidates and customer content
I pull raw input from multiple sources: community forum threads, support cases, Slack channels, idea submissions, customer calls. I feed that into Claude with context on what release cycle I'm in and what problems I'm trying to solve. It synthesizes patterns, surfaces recurring themes, and outputs a ranked list of backlog candidates — each one with a short rationale tied back to the source signal. The same synthesis also generates a first draft of customer-facing content: how-to guides, release notes, and enablement material that explains the feature in language customers actually use, not product jargon.
Product specs: epics, stories, and announcements
This is the AI product spec workflow that saves me half a day, every sprint. Once I know what I'm building, Claude helps me go from rough idea to structured spec. I describe the problem, the customer, and the outcome I want. It drafts the epic, breaks it into stories with acceptance criteria, and generates a first cut of the release announcement. I review and edit — the judgment is mine — but I'm no longer starting from a blank page. A spec that used to take half a day now takes an hour, and it's more consistent because the structure is always the same.
What actually changed
The obvious win is speed. But the less obvious one is quality — when you're not spending energy on structure and formatting, you can spend it on the actual thinking. My specs are sharper. My backlog is better prioritized. And my customer content is closer to what customers need because it's built from their own language, not mine.
What I'd tell other PMs
Don't start with a tool. Start with a friction point. What's the thing you do every week that feels like busywork but you can't skip? That's your first workflow.
The system doesn't need to be complex. Mine runs entirely through Claude and Notion — no elaborate integrations, no code. Just clear prompts, consistent structure, and the discipline to actually use it.
The goal isn't to automate your job. It's to protect your time for the parts that actually need you.