If you’ve ever shipped an AI feature that looked great in testing - only to watch it behave unpredictably in production - you’re not alone.
In this episode of IT Visionaries, host Chris Brandt talks with Lawrence Jones, Founding Engineer at incident.io, about the critical gap between AI that demos well and AI that works under pressure. Lawrence shares how his team designs tools that help engineers respond faster, learn from failure, and build systems that don’t crumble when it counts.
CHAPTERS / KEY MOMENTS
00:00 - AI Chaos & The Mike Tyson Rule
00:58 - Meet Lawrence Jones of Incident.io
03:14 - From FinTech Outages to Incident Response
06:22 - The Biggest Mistake in Incident Management
09:08 - Training for Chaos: Game Day Simulations
10:31 - Inside the AI SRE System
13:01 - What SRE Really Means
16:23 - From Prototype to Production AI
20:27 - Keeping Up with AI’s Rapid Evolution
22:50 - Understanding Vector Databases & Embeddings
28:34 - The Architecture Problem: Chaining Prompts at Scale
36:11 - Measuring AI Performance & Reliability
44:02 - The Future of SRE Meets AI
52:10 - Lessons from Real Incidents
56:42 - Final Thoughts: Building AI That Works