Don’t Become Case 15: Why AI Projects Fail Even When The Systems Are Right

AI Failures rarely start with the model. They start with a decision nobody was equipped to make, or a warning nobody was empowered to escalate. This isn’t just about failed software delivery, it’s about the decisions that quietly fail around it, long before anyone notices.

Drawing on fourteen forensic case studies, including Post Office Horizon, Boeing MCAS, Knight Capital and the Dutch child benefits scandal (Toeslagenaffaire), this talk unpacks the “interpretation gap”, the space between what a system reports and what leadership understands. Andrew will walk through recognisable failure patterns and show how they show up in AI-driven decisions well before the technical postmortem begins.

The talk closes with a practical framework for closing the gap before it becomes your organisations next case study. Expect real cases, hard questions, and time for discussion.