The PM weekly review is where you either gain control of your week or let it control you. The old way is 2 to 3 hours on Monday morning digging through Slack, Jira, analytics dashboards, and email to figure out what happened and what should happen next. The new way is 30 minutes: let AI pull the data and prepare a brief, then spend your remaining time on what only a PM can do, deciding what ships, unblocking the team, and talking to customers. This is the exact system I use and have refined with a dozen product teams.
Phase 1: gather, 10 minutes
Aggregate five data sources: engineering from GitHub, Jira, or Linear, product metrics, customer feedback from support, sales from Salesforce, and marketing. Then run a brief prompt in Claude that produces a headline, what shipped, what has momentum, what is at risk, two or three key insights, and this week's external context. Format: scannable bullets, max 400 words. It takes 2 to 3 minutes and replaces skimming 50 Slack messages. Pro tip: include last week's brief so the AI surfaces week-over-week trends automatically.
Phase 2: triage, 10 minutes
Now you decide. Ask the brief five questions. Is anything broken that was not yesterday? Did we learn something that changes the roadmap? Are we hitting our goals? Is anyone blocked waiting for me? Did we miss something? A yes routes to an action: a ten-minute sync with the owner, a flag for roadmap review, an escalation if it is a trend rather than one-week noise, or a decision you make before anything else today. Everything else gets noted but does not change what you are shipping.
Phase 3: communicate, 10 minutes
Generate a three-minute standup opener that leads with the headline, explains how it changes what you are shipping, calls out blockers, and ends with what you are focused on this week and why. Make it conversational and honest. Paste it into your Monday standup. You are done with the review.
Catch the slow trends
Every month, on the last Monday, run a four-week trend prompt: paste all four weekly briefs and ask what patterns will not show up in a single week. This catches slow-moving problems that look fine in isolation, like a metric down 2 percent four weeks running.
A few mistakes to avoid. The brief is input to your decision, not the decision, so keep synthesis and judgment separate. Do not over-automate and lose signal. And do not treat the review like a report. Thirty minutes tops. If you do not want to build automation yet, just ask your team every Friday to post three sentences each, engineering on what shipped and what is blocked, sales on deals and objections, support on the top three issues, and paste that in. Start with the prompt, not the plumbing. Block Monday 8:00 to 8:30 this week and run all three phases.