The User-Feedback Distiller
Separates what users said from what we should build — the synthesis layer most product teams skip and then regret.
What it does
Weekly, this agent ingests user feedback from every channel (sales calls, support tickets, product analytics events, in-app surveys) and produces two distinct outputs: (1) a "what users literally said" report with verbatim quotes ranked by frequency, and (2) a "what this means we should consider building" memo with the strategic interpretation separated from the raw data. Stops PMs from confusing customer requests with the underlying jobs-to-be-done.
The Manus prompt
Copy this prompt and paste it into Manus AI to run the agent.
Every Monday at 7am, pull the last 7 days of: customer-facing support tickets in Zendesk, sales call notes in our CRM, in-app survey responses, and product analytics anomalies (any feature with usage drop >25% or spike >50%). Output two documents: (1) "User signals — verbatim" — direct quotes and event data, ranked by frequency, with no interpretation; (2) "User signals — interpreted" — the strategic synthesis showing which underlying job-to-be-done each signal points to, what we'd build if true, and the risk if we're wrong. Save both to a Notion folder, email me when ready.
Tools required
LinearJiraZendeskNotionSalesforce
Best for
ProductEngineering