Multi-Agent Interviewer System for Expert Personas

I wrote about how we built a multi-agent interviewer system at 99Ravens to capture tacit expertise and package it into first-person system prompts. The key is pairing an interviewing agent with an observing agent that tracks coverage, flags generic answers, and proposes targeted follow-ups so the conversation stays sharp and peer-like—even over long sessions.

Read the full breakdown and architecture here: Your experts won’t train your AI—you have to interview them.