AI technical due diligence service for investors and acquirers evaluating automation claims
A diligence framework that stress-tests an AI startup's automation percentage before it reaches a term sheet: arithmetic on workload vs. fully loaded labor, a named automation boundary with a straight-through rate, evaluation credibility checks, and finally unit economics per accepted autonomous task.
From AI Engineer — What Technical Due Diligence Finds When the Deck Says “70% Automated” — Soukaïna Abelhad at 00:00
Problem: Automation numbers are often genuine measurements carried into conclusions they cannot support — a percentage with no denominator hides whether the work is assistance or labor removal, and automated graders can be far more optimistic than the experts who accept the output.
For: Investors, acquirers and buyers who are shown automation metrics like '70% automated' or '55.8% faster' in a deck and must decide whether to believe them.
Examples
- A controlled experiment with 95 professional developers showed 55.8% faster, but only on one JavaScript HTTP server built from a template against 12 visible tests.
Other takes on AI evaluation and benchmarking
- An independent evaluation company for foundation models and agents
- Per-agent tracking and evaluation tables for ROI reporting
- Lint agent skills so their code examples are real, runnable, and schema-accurate — including anti-gaming rules
- Context engine simulator for A/B testing whether pre-gathered context improves your coding agents