Startup using CrewAI for automated due diligence
“Runs complete startup due diligence analysis in 90 seconds, producing structured investment reports with scored verdicts (INVEST / CONDITIONAL / WATCH / PASS) and machine-readable JSON output.”
solo-developer
Why they built it
Part of a 15-day hands-on series to learn agentic AI progressively, from single agents to multi-agent systems with APIs, memory, and delegation, specifically for fast startup investment evaluation.
What worked
Hierarchical delegation where each agent passes output as context to the next; JSON output integrates directly into tools like Notion/Airtable without copy-paste.
What broke or was painful
Handling LLM output variability required a two-pass JSON parser for cases where LLMs drop markdown fences; challenges in structuring task descriptions for good agent reasoning and designing for failure modes.
The result
Runs complete startup due diligence analysis in 90 seconds, producing structured investment reports with scored verdicts (INVEST / CONDITIONAL / WATCH / PASS) and machine-readable JSON output.