The conditions for failure
Ambiguous ownership, poor incentives, weak evidence, and decisions made too far from the people affected.
Narrative nonfiction · 2026
When the Model
Was Wrong.
Stories from the front lines of system failure.
The book
System failures are rarely only technical.
A case-led account of what happens when automated systems meet institutional blind spots.
When the Model Was Wrong reconstructs consequential failures and follows responsibility beyond the model itself: into incentives, data, procurement, organizational design, oversight, and the decisions people made before and after deployment.
Written for executives, boards, practitioners, students, and readers who want to understand not only how systems fail, but why institutions repeatedly fail to see it coming.
Reader response
Reader testimonial
“I think this is the most readable book about AI governance you could ever hope to find.”
“This book is the closest to a comprehensive technical blueprint for governing complex AI systems that I have found.”
“I’ve not read anything that explains and precisely points out the challenges with complex, interconnected, and automated decision making systems like this book.”
The argument
The model is only one part of the story.
The moment the system breaks
What the failure looked like in practice, who saw it first, and why intervention came too late.
The institutional response
How organizations explain, investigate, contain, and sometimes misunderstand what happened.
What responsible leaders change
The governance, design, oversight, and escalation choices that make a different outcome possible.
For readers
Cases worth remembering before the next decision.
- Executives and board members
- Technology and product leaders
- Risk, legal, privacy, and security teams
- Faculty, students, and public-policy audiences
- Anyone responsible for automated decisions