AI trust in regulated industries isn't a compliance checkbox — it's an operational discipline you build into every iteration. Abhijeet Gulati, head of AI at Mitchell (an Enlyte company), has spent nine years shipping AI into property and casualty insurance claims management, where accuracy is non-negotiable and a wrong estimate has real financial consequences for real people.

In this conversation, Abhijeet introduces the velocity-veracity paradox — the tension between time to market and time to trust — and breaks down how his team resolves it. You'll hear how Mitchell decides when AI is production-ready vs. when to pause, why bias must be operationalized as a routine rather than treated as a one-time audit, and how human-in-the-loop design works in practice when AI confidence meets uncertainty.

This is the final episode of our three-part series on AI as an operations force multiplier: EP32: AI Didn't Replace These Workers — It Gave Them Their Mission Back — https://youtu.be/ziCYZtWNzps EP33: How HCA Turns Clinical Notes Into Intelligence — https://youtu.be/XAGaIPnuUhY

Book an Insight Prism workshop because you'll get a structured framework to identify where AI can create operational intelligence from your existing data: https://www.insight.com/en_US/what-we-do/methodology/insight-prism.html

Subscribe to Insight On for new episodes every week.

#AItrust #RegulatedAI #ResponsibleAI #InsightOn

Chapters

• 00:00 — Welcome and introduction

• 01:57 — What Mitchell does in claims management

• 03:23 — How AI powers the insurance claims experience

• 05:22 — The velocity-veracity paradox explained

• 08:00 — Building and maintaining customer trust with AI

• 10:18 — Build vs. buy decisions for regulated AI

• 11:53 — What generative AI makes possible now

• 13:26 — Framework for deciding when AI is ready to ship

• 15:29 — Operationalizing bias as a continuous routine

• 18:09 — The one question every AI leader should ask

• 20:35 — Closing thoughts