AI-Powered Deal Origination

The challenge

The fund was spending a disproportionate amount of time on manual deal screening. Their existing origination platform was expensive and rigid — it could not be configured to their specific mandate criteria around sector, geography, revenue profile, and deal stage. The alternative was manual screening by the investment team, which was slow and pulled people away from higher-value work.

They needed a system that could process opportunities at volume, score them against their exact mandate, and present the investment team with only the deals worth looking at — without the cost of an enterprise origination platform.

What we built

We built a custom AI origination engine integrated directly into their CRM. The system uses a multi-LLM approach — Claude, GPT-4.5, and Gemini — to classify and score each opportunity against the fund's mandate. A rules-based layer on top handles the deterministic elements: sector filters, geography, revenue thresholds, and deal type.

• Multi-LLM classification (Claude, GPT-4.5, Gemini) with ensemble scoring
• Automated priority scoring by sector, geography, revenue profile, and deal type
• Full mandate parameterisation — criteria configured to the fund, not a generic template
• Fireflies AI integration for deal call transcription, auto-saved to Notion with follow-up email drafts
• CRM integration — opportunities flow in, scored, and pushed back with deal memo drafts

The results

The system processed over 10,000 investment opportunities in its first 24 hours of operation. The investment team went from spending hours per week on manual screening to reviewing a curated shortlist of high-priority deals each morning.

Classification accuracy came in at over 95%, and the total cost of the system was approximately 25% of their previous origination platform. Each member of the investment team saved over 10 hours per month — time redirected to relationship building and deal execution.