
Episode #48
90% Adoption, 8,500 Agents, and What We'd Change If We Started Over | EP48
More than two years ago, Insight made a bold call: give every one of our 14,000 teammates access to AI tools and tell them to use it everywhere. This episode is the full story of what that journey built, broke, and taught us — the 90% adoption rate, the 8,500 agents, the 49,000 hours of meetings summarized in a single month, and the things our CTO Mark Hughes would do differently if we started over. In this conversation, Mark walks through the real arc of our AI transformation — from the early "use it everywhere" mandate to the hard pivot toward cost discipline, governance, and smarter deployment. You'll hear why solving the AI problem is really about solving the data problem first, how a 10X overnight cost increase forced a company-wide rethink of token spending, and what "human in the loop" actually means when AI confidently makes things up. Whether you're building your AI operating model, wrestling with ROI conversations, or trying to get your organization past the pilot stage, this episode gives you a practitioner's view of what AI adoption actually takes — data readiness, governance, change management, and all. Get Insight's AI and data solutions because it will help you move past pilots and into AI that actually sticks : insight.com/data-and-ai Also worth exploring: Radius AI: insight.com/radius-ai AI Flight Academy: insight.com/ai-flight-academy Enterprise AI Hub: insight.com/enterprise-ai-hub Getting AI Unstuck — From Strategy to Scale: insight.com/getting-ai-unstuck Subscribe to Insight On for more conversations with practitioners building real AI programs inside real organizations. #AIAdoption #AITransformation #GenerativeAI #AIStrategy Chapters 00:00 — Welcome and intro 02:41 — How our earliest AI adopters got started 03:09 — The AI ROI question nobody has a clean answer to 04:06 — When token costs jumped 10X overnight 07:21 — What 49,000 hours of meeting summaries actually means 09:13 — Managing 8,500 agents and the duplication problem 14:00 — Solving the AI problem by solving the data problem first 18:44 — What Flight Academy did for AI training at scale 21:19 — Human in the loop — and why we're thinking about it wrong 29:11 — The Horizon platform: from chat interface to agentic orchestration



