Board of Advisors

Board of Advisors is playful, but technically concrete: several agent personas receive a problem, discuss it, and the result can be processed into JSON, a story, or audio.

A single LLM answer is often too smooth. For decisions and messy problems, multiple points of view, critique, and narrative can be more useful.

A few strongly different perspectives can break a flat LLM answer faster than adding another paragraph to a single chatbot prompt.

• Persona agents with distinct conversational styles.
• Conversation scenarios and a user proxy to introduce a problem.
• Output processing into JSON/story format.
• Experimentation with a TTS/storytelling pipeline.

AutoGen manages the agent dialogue, while OpenAI/GPT generates responses aligned with each persona. The output can be processed further into story or audio.

Python, AutoGen, OpenAI API, Streamlit/GPT, and a helper repo for turning JSON into a story.

• Several specialized perspectives instead of one generic assistant.
• Structured output so the conversation can be reused downstream.
• Learning experiment focused on product direction, not a final SaaS.

Working experiment with AutoGen, GPT, personas, structured output, and a storytelling/TTS pipeline.

Turn the experiment into a more practical advisory process for founders or product teams.