Sales and Product Intelligence Engine
The Challenge
During my time as a Product Manager, valuable customer feedback was usually buried in 30–40 minute Zoom recordings done by sales. Because summarizing these calls was a manual task—requiring PM or Sales to re-watch, synthesize, and document—only about 25% of calls were ever logged.
When documentation did happen, it was often days late. This left leadership and the product team making roadmap decisions without a complete picture of what users voiced during those calls. I needed a way to capture insights from every call without the 1.5-hour manual rewatch-and-synthesize cycle per recording.
The Solution & How It Works
I built an automated pipeline using n8n that triggers the moment a Zoom recording is finalized and uploaded into the pipeline.
- —The Workflow: The system pulls the recording, generates a transcript, and passes it to an AI model with a specific prompt designed to extract unmet user needs and feature requests.
- —The Output: Summaries of each call are automatically pushed to a Notion "feedback database" space and then shared in a dedicated "insights" Slack channel.
- —The Shift: This replaced 90 minutes of manual rewatch-synthesize-document-share with an instant Notion document for every call.
Execution
I used n8n to orchestrate the data flow between Zoom recording, transcription, Notion, and Slack.
The core of the project was prompt engineering. To ensure the AI didn't produce noise or hallucinations, I implemented a simple evaluation process: I audited automated summaries against my own manual notes for about 10 calls to refine the prompt. This iterative testing ensured the output was accurate and focused on actionable product data rather than general conversation.
The Impact
- —100% Documentation: Every sales and discovery call became a structured insight doc automatically.
- —Time Efficiency: Saved PM and Sales teams 1–1.5 hours per call.
- —Data-Driven Roadmap: The product roadmap was updated without missing any customer data voiced during calls.