Most companies are asking the wrong question about AI — and a Stanford researcher's field work shows exactly where the returns break down. Melissa Valentine, professor at Stanford University, senior fellow at the Stanford Institute for Human-Centered AI, and co-author of Flash Teams, makes the case that AI workflow integration — not better prompting — is what separates real ROI from endless experimentation.

In this conversation, Valentine shares findings from her research on why high-impact AI users think like product managers, why the org chart isn't dying but needs to be understood differently in an AI-powered organization, and why the efficiency narrative around AI may be leading organizations straight into the Turing Trap.

Subscribe to Insight On for conversations with researchers, practitioners, and business leaders on the technology decisions that shape outcomes.

📖 Read Melissa's research:

To Drive AI Adoption, Build Your Team's Product Management Skills: https://hbr.org/2026/02/to-drive-ai-adoption-build-your-teams-product-management-skills

How to Make Enterprise Gen AI Work: https://hbr.org/2025/09/how-to-make-enterprise-gen-ai-work

Flash Teams (book): https://www.amazon.com/Flash-Teams-Leading-AI-Enhanced-Demand/dp/0262049848

Jump to…

00:00 — Welcome and guest introduction

03:54 — The real reason companies aren't seeing AI ROI

05:18 — The copy-paste trap and integrated workflow problem

09:12 — The product manager mindset for AI adoption

14:18 — Overcoming resistance to workflow change

18:08 — Levels of AI maturity across organizations

25:29 — Is the org chart dying or just changing?

28:29 — Why one-to-one agent replacement is the wrong model

32:11 — What a working agentic system actually looks like

33:36 — Cognitive load and the hidden cost of AI-accelerated work

35:38 — How to bring a CFO along on AI transformation

Stop Replacing Humans With Agents. It Doesn't Work That Way | EP24