AI Due Diligence for Investors

Every pitch deck claims AI. Most of it is a thin API wrapper around GPT-4 with no defensibility, no custom training, and no plan for when OpenAI raises prices by 3x. Investors need to know the difference between genuine AI capability and demo-ware before they write a cheque.

We provide independent technical assessment that separates real AI moats from marketing slides.

What We Assess

Model and Algorithm Evaluation — Is the AI actually doing something proprietary, or is it a prompt template? We evaluate whether the company has genuine technical differentiation — custom fine-tuning, proprietary data advantages, novel architectures, or unique training pipelines.

Infrastructure Maturity — Can this system handle 100x the current load? We assess architecture, deployment practices, monitoring, failover, and cost scaling. A system that works for 100 users is not the same as one that works for 100,000.

Team Assessment — Does the engineering team have the skills to maintain and improve the AI systems? We evaluate team composition, technical depth, and knowledge concentration risk (the "what happens if one person leaves" question).

Cost Structure Analysis — What are the real unit economics? We model inference costs at current and projected scale, identify hidden costs (data labelling, retraining, monitoring), and assess whether the company can maintain margins as it grows.

Security and Compliance Readiness — Is the AI system ready for enterprise customers? We assess data handling, access controls, audit capabilities, and compliance readiness against SOC2, ISO 27001, and sector-specific requirements.

The Deliverable

A written report with clear risk ratings (Green / Amber / Red) across each assessment dimension, specific findings with evidence, and actionable recommendations. We present findings to the investment committee and answer technical questions.