What Technical Due Diligence Finds When the Deck Says “70% Automated” — Soukaïna Abelhad

A pitch deck says a workflow is 70% automated. Before accepting the number, Soukaïna Abelhad asks what it counts, what human work remains and who checked the result. Her first diligence test needs no model expertise: compare workload growth with the fully loaded labor required to deliver it. More output with proportionally more reviewers, trainers, escalations and rework can describe assistance rather than labor removal. A percentage without a clear denominator obscures that distinction, especially when the work called automated still depends on people to complete or correct it. The second test defines the level of automation inside a named workflow. AI attributed work and an autonomously completed task are different claims; the straight through completion rate needs a boundary that a buyer can inspect. The third test examines evaluation. A model judging another model's output is useful evidence only within its limits, and agreement with an automated rubric can diverge from a domain expert's assessment. Abelhad turns those checks into an economic question: what is the cost per accepted autonomous task after model costs, human review, escalation and rework? Failures belong in the denominator and the cost account. Public examples and a brief anonymous evaluation illustrate the framework rather than establish a general automation rate or a causal explanation for a company's failure. The takeaway is a practical diligence conversation that connects the deck's language to observable work, credible evaluation and the operating cost behind the claim. Speaker info: - https://www.linkedin.com/in/soukaina-abelhad/ Timestamps: 00:00 - What an automation percentage leaves out 0:55 - Check one: workload against human labor 2:13 - Check two: define the automation boundary 3:05 - Check three: evaluation and expert review