What Technical Due Diligence Finds When the Deck Says “70% Automated” — Soukaïna Abelhad
Technical audits reveal AI automation claims typically describe assistance, not labor elimination
“A dangerous number is not always fabricated. This is often a true measurement that leads to a conclusion that it cannot confirm.”
Soukaïna Abelhad presents a framework for auditing inflated AI automation claims, showing that figures like '70% automated' rarely mean end-to-end autonomy. Faros telemetry across 10,000 developers shows AI-heavy teams saw 98% more merge requests and 91% longer review times, suggesting AI expands workload volume rather than eliminating human labor. She introduces a definitional ladder — from AI-drafted to fully autonomous — to force precise claims during technical due diligence.