Why Industrial AI Fails
What the 5% Do Differently, and What They Build First
A book for the people who run plants, utilities, and operations, by Robert Hylton. Setpoint Press, January 2027.
The five. In 2025, an MIT team reported that roughly 95 percent of enterprise generative-AI initiatives delivered no measurable return. This book is about what the other five build first.
Data + limit = KPI
347° means nothing on its own. It means something when someone wrote down that over 340 is bad.
One tank, six names
TK_401, T-401A, North Feed Tank, asset_88213. The model was fine. Nobody could agree what it was looking at.
The 2:49 a.m. alert
A system that brings the answer to you, at the moment someone is about to make the wrong call.
Before you fund the next AI pilot, answer five questions.
The AI Readiness Check: five questions from the book to answer with your team.
About Robert
Robert Hylton is a co-founder of Transpara, an industrial software company, and has spent his career in the software industry.
now
Co-founder, Transpara
Industrial software for the people who run plants, utilities and operations.
before
Microsoft
Director of the Server & Tools division for Asia Pacific and Greater China. Led product management teams in Redmond, and was Microsoft’s global lead for supply chain, B2B marketplaces and supplier enablement.
writing
Two books
Why Industrial AI Fails (Setpoint Press, 2027). Solutions for Supplier Empowerment (Microsoft Press, 2002), principal author.
One email, on launch day.
No newsletter. Where to buy it, and nothing else.
