Product Management AI Agent

Inspiration

We are an early stage AI infra company focused on connecting enterprise data sources to AI agents. In the last couple weeks, we finished our proof of concept of this context retrieval platform and were looking for a good opportunity to build an AI agent on top.

What it does

First the agent comes up with 6 questions/problem spaces it wants to investigate in the company. It then finds the answers to those questions using data from across the enterprise using the Mimo API. Based on the responses, it comes up with another batch of 6 questions to investigate further. Then, using everything it's found as notes, it creates a list of new features with rationale for each one and passes it back to the user. The user chooses which feature they like best. The AI then writes a PRD and creates the relevant Linear tickets to get the feature built.

How we built it

We built our own framework for developing AI agents. One of the tools the agent has access to is the Mimo API which is an enterprise context retrieval platform.

Challenges we ran into

We found it impossible to use LangChain, AutoGPT, or even BabyAGI meaningfully and had to rewrite much of the underlying infrastructure ourselves.

What's next for Product Management AI Agent

We are considering spinning off the agent framework that we developed in this hackathon as an open source project to compete with AutoGPT and BabyAGI. Also, we will release Mimo for developer access soon.

Built With

• linear
• mimo
• next.js
• openai
• tailwindcss
• typescript
• vercel