Real AI

An AI assistant at a financial technology company took over the work of roughly 700 customer service representatives. A little over a year later, after service quality slipped, the company began hiring human agents again, though the company describes the new hiring as a small pilot. The gap between what AI can do in a demo and what it does inside real work is the subject of this book. The Real AI ignores the two conversations that dominate the headlines: the frontier labs racing toward ever-larger models and the debate over whether AI will deliver abundance or catastrophe. Instead, Mark DeSantis and Anand S. Rao turn to the operational middle ground, where AI either earns its place in a workflow or quietly fails. Drawing on cases from UPS, HSBC, GE, Palantir, India's Aadhaar, and Michigan's MiDAS unemployment system, along with thirteen interviews with the people who build, buy, fund, govern, and clean up after AI, they show that models are the easy part. What matters is the infrastructure around them: learning loops, trust management, organizational absorption, and the accumulation of what the authors call cognitive capital, the durable store of experience that lets a system get better with use rather than decay. Organizations that build cognitive capital improve faster than those that do not, and that learning advantage is what the book shows leaders how to create. The book moves in three parts. Building the Real AI explains how working systems are engineered, why expertise alone is not enough, and why failure, risk, and harm are three different things. Embedding the Real AI examines how adoption unfolds in regulated industries, operations-driven businesses, the public sector, and capital markets, each with its own judge. Shaping the Real AI turns to geopolitics, national security, designing for difference, and global governance. For executives, founders, and policymakers, this is a field guide to making AI real in the work you do, the organizations you lead, and the society you shape.