🪄Using Claude for PM Work
The day-to-day Claude workflows that compound a PM's leverage. From PRDs to user research synthesis to roadmap analysis.
Using Claude (or GPT-5, or Gemini) daily for real PM work is the fastest way to build AI literacy AND boost your own productivity. PMs who do this ship 2-3x more strategic output than those who don't.
The high-leverage Claude workflows for PMs: drafting PRDs from bullet points, synthesizing user research, analyzing data and SQL, exploring strategy framings, prepping for meetings, writing exec updates. The pattern is the same: PM provides judgment + context; Claude provides typing + first-draft structure.
The dozen workflows that compound
1. PRD drafting. Paste bullet points + context. "Draft a PRD with TL;DR, problem, solution, metrics, risks." Iterate.
2. User research synthesis. Paste 5 interview transcripts. "Identify the 5 most common patterns. Quote verbatim. Tag each as opportunity / pain / solution candidate."
3. SQL writing. "Write a SQL query to get D7 retention by signup cohort by acquisition source." Validates before running.
4. Data analysis. Paste a CSV. "What patterns do you see? What's the most interesting cohort?"
5. Strategy framings. "I'm trying to decide between A and B for next quarter. Steelman both. What am I missing?"
6. Meeting prep. "I'm meeting [person] at [company]. Here's their background. What should I prepare?"
7. Exec updates. "Here are the team's accomplishments this week. Write a 4-sentence update for our CPO."
8. Customer email drafts. Pattern of the user's complaint + context. "Draft a response."
9. Job description writing. Role + must-haves. "Draft a JD that's specific and not generic."
10. Code review. Paste a PR. "What's missing? What edge cases?"
11. Vibe coding small tools. "Build a script that pulls our daily metrics into a Slack message."
12. Self-review of writing. Paste your draft. "Critique this. What's weak?"
The patterns that work
- Lots of context, then ask. Don't ask 'how do I write a PRD' — paste the bullets and ask 'turn this into a PRD.'
- Iterate, don't take the first draft. Always ask 'critique this' on the first output.
- Use it for the typing, not the thinking. Judgment is yours; the words are Claude's.
- Project mode for ongoing work. Persistent context across many conversations.
What it's bad at (in 2026, may change)
- Domain-specific judgment without context. Always feed it the context.
- Knowing your stakeholders. Generic exec updates fall flat.
- Spotting subtle product taste issues. A great PM still has better taste than the model on nuanced design questions.
The daily PM workflow
Many senior PMs in 2026 have Claude open in a side window all day:
- Morning: review the day's tasks with Claude. Surface anything urgent.
- During PRDs: draft sections, get Claude to critique, revise.
- During analysis: paste data, get patterns surfaced.
- During meetings: Claude open to summarize quickly.
- End of day: ask Claude to summarize what shipped, draft tomorrow's plan.
The compounding effect: 30-50% productivity boost for written work, with no quality loss when done well.
Real-world examples
PMs at Anthropic obviously use Claude heavily. The patterns they share — projects mode, MCP for tool integration, careful prompt design for repeatable tasks — are seeping into the broader PM community as best practices.
Go deeper — recommended reading
Interview questions (1)
Q1Walk me through how you use Claude or ChatGPT day-to-day as a PM.ai-pmmid▼
I have Claude open most of the day. Specific workflows:
- PRDs. I draft bullets + context, Claude turns into prose. I revise; that's the actual judgment work.
- Research synthesis. Paste interview transcripts, get patterns surfaced. Saves hours of manual coding.
- SQL. Write queries faster than I would alone. Validates before I run on production data.
- Exec updates. Bullet what shipped, get a 4-sentence narrative back.
- Strategy framings. "Steelman the case against this." Forces me to consider what I'm missing.
- Critique of my own writing. Paste a draft, ask 'what's weak?'
The pattern: Claude provides typing and first-draft structure; I provide judgment and context. Compound effect is ~30-50% more strategic output per week with no quality loss.
What I don't use it for: subtle product taste decisions (still need my own gut), conversations that require knowing my stakeholders (generic exec updates fall flat), or anything requiring a deep model of our specific market and customers.