The Conversation

What Does It Take to Move AI From Pilot to Production?

Generative AI can write a 50-page report in seconds. The harder question is whether anyone can trust what's inside it.

In this 60-minute conversation, three practitioners cut through the hype and talk plainly about the gap between content creation and production-grade AI.

The through-line is simple. AI gives every employee superpowers. But superpowers without verification create new failure modes that businesses, regulators, and auditors are not yet equipped to catch. Closing that gap is what "trust at scale" actually means.

Four ideas worth your hour.

The conversation moves through four interlocking ideas that define what production-grade enterprise AI actually requires.

icon_the_moat

The Moat

Why tacit knowledge is the moat. Every company has a "Colonel Sanders recipe" the foundation models do not have, and the question is how to feed it into the AI harness without losing it.

icon_the_ai_harness

The AI Harness

Why the AI harness matters more than the model. Models have largely commoditized. The competitive edge is now the substrate of process, context, and data each company builds around them.

icon_model_portability

Model Portability

Why model portability is now a strategic imperative, not a procurement preference, and what changes when you switch foundation models the way you change cars.

icon_building_trust

Building Trust

The shift from generation to verification, and why companies that engineer trust into their AI systems will operate at a level their competitors cannot reach.

A practitioner's view of enterprise AI.

Why business AI is harder to validate than developer AI, and what that means for adoption inside regulated industries.

How Resolution Life regression tested AI outputs agains 16 months of actuaral calcuations.

What "AI native" looks like at the individual level, and why adoption has become the bottleneck, not capability.

The "AI harness".  Why companies that own their substrate of content, process, and institutional knowledge are pulling ahead.

Why model portability is now a strategic imperative, not a procurement preference.

The shift from generation to verification, and what financial institutions are actually building underneath their AI deployments.

Meet the Speakers

Peter Roschke

Peter Roschke

Peter is the Co-founder and Chief Technology Officer of Coherent. He leads technology strategy and platform innovation, where the team has built the operating system for enterprise Excel. His work focuses on how financial institutions gain visibility, control, and confidence over the spreadsheet logic that runs the business. 

Purav Desai

Purav Desai

Purav leads AI strategy at Resolution Life, a leading multinational life insurer and asset manager. He has driven the firm's shift from generative AI experimentation to verified production deployment across 1,200 products and a multi-decade actuarial estate.

David Simmons

David Simmons

David leads the strategy and data science function at Resolution Life. He hosts this conversation, drawing on his work bridging actuarial discipline, modern data infrastructure, and the operational realities of moving AI into a regulated business.