Product Mate

From user meetings to measurable impact in minutes, not days.

Comment

Product Mate: The Self-Improving AI Notetaker for Product Managers

From user meetings to measurable impact in minutes, not days.

Problem

Product managers spend hours turning user feedback into action — and most insights die in meeting notes or spreadsheets.

Pain Points:

  • Lost insights: 50–80% of qualitative feedback never becomes actionable.
  • Manual work: PMs rewatch interviews, summarize notes, and manually write Jira tickets.
  • Redundancy: Teams repeatedly create tickets for issues already logged.
  • Slow iteration: Valuable user pain points take weeks to surface and resolve.

Solution (How Product Mate Works)

1. Capture & Transcribe

Our desktop app records Zoom/Meet sessions and generates complete transcripts.

2. Extract & Learn

  • LLM extracts key insights.
  • A self-improving reinforcement learning agent iterates to generate the most personalized, factual, and actionable action items.
Type Description
Factual Uses Tavily Search API + searches existing GitHub issues & Jira tickets to confirm that concerns haven’t already been addressed.
Actionable Avoids vague observations; focuses on concrete steps.
Personalized Prioritizes each action item (High / Medium / Low).

Before creating new tickets, Product Mate runs cosine similarity checks on Jira and GitHub issues to ensure no duplicates.

3. Delegate & Close the Loop

Automatically creates and assigns prioritized Jira tickets to the correct team — Product, Engineering, Design, or Marketing.

The Self-Improvement Loop

Product Mate’s agent learns from every meeting:

  1. Observation — Transcribes & summarizes calls.
  2. Evaluation — Scores action items using an RL-based reward function for factuality, actionability, and priority.
  3. Reflection — Reviews past outcomes (e.g., whether prior tickets were resolved).
  4. Adaptation — Refines its own generation strategy for next time.

Future Vision

If given more time, we plan to build:

  • Interactive Dashboard: Visualize themes, clusters, and recurring user pain points.
  • Cross-Team Insight Matching: Identify overlapping feedback across interviews.
  • Adaptive Prioritization: Continuously fine-tune recommendations based on historical impact.

Team

Amogh Maheshwari • Sohan Dillikar • Savir Dillikar

Built With

Share this project:

Updates