Menu
AI Application Development · Voice AI · Generative AI · Web App Development · Multi-Agent Systems
For a leading venture capital firm, the early stage of founder screening is among the most time-consuming parts of the investment process. Every week, dozens of founders send in pitches, each requiring a structured conversation, a set of consistent diligence questions, and a clear written summary for the investment team to act on.
We built Pitch Fabrice to handle that entire first stage automatically. It is a voice-first AI platform that conducts real-time conversational interviews with founders, asks the right questions based on their stage and business model, and generates an investor-ready debrief report after every session, ready for internal review without any manual note-taking or follow-up calls.
Early-stage screening is valuable but resource-intensive. The team needed to process more deal flow without sacrificing the consistency or depth of the diligence process.
Early and growth-stage startups need different questions. The system had to adapt intelligently to each founder’s context.
Founders needed a genuine conversation, not a form. A mechanical experience would undermine trust and shallow the answers.
The team needed a clean, structured debrief after each interview, with no one writing it manually.
Interviews get interrupted or rescheduled, so sessions had to pause and resume without losing context.
We designed Pitch Fabrice as a multi-agent voice platform, with a supervisor AI orchestrating a pipeline of specialised agents, each built for a specific founder stage: Early Stage, Growth Stage, Confirmation, and Pitch Deck Analysis.
When a founder joins, the system identifies their stage and routes them to the right agent, which runs a structured conversational interview across background, business model, market size, traction, and fundraising terms. The conversation flows naturally through OpenAI’s Realtime API. Before the interview, founders upload their pitch deck, which vision AI reads and cross-references so the agent already knows the basics.
After the session, the platform automatically generates a structured debrief covering company summary, fundraising instrument classification (SAFE, convertible note, priced equity), traction, and stage-specific insights. Sessions are fully persistent on MongoDB, so founders can pause and resume at any point.
Natural voice conversations across background, business model, market, traction, and fundraising terms.
A supervisor AI routes each founder to the right question set by stage, using the OpenAI Agents SDK.
Vision AI extracts and cross-references the uploaded deck so questions are sharper and more relevant.
Investor-ready reports generated automatically after each session, no manual effort.
MongoDB-backed state with full context replay, so interviews pause and resume cleanly.
Next.js (TypeScript), OpenAI Realtime API, OpenAI Agents SDK, MongoDB, AWS App Runner, Docker.
Pitch Fabrice significantly reduces the manual effort of initial founder screening, so the VC team can process more deal flow with consistent, structured data from every conversation. What once took hours of calls and note-taking now happens automatically, freeing the team for the decisions that need human judgment.
© 2026 Zimozi Solutions
