Jeetu Patel said the quiet part at Webexone 2026: “Most people a couple layers below get afraid their jobs will be done, so we made it easy for our people.” @nyike's write-up of the event makes culture the first job for CIOs on AI adoption. I agree, and the order Cisco followed is the part most enterprises get wrong. Arvind Jain, @Glean told the room he isn’t seeing enough top-down AI transformation. I hear the same thing in client conversations every week. Employees build agents on their own, a lot of what they build is low value, and sprawl follows. The CIO then gets asked to clean it up and show ROI on work nobody scoped. @jpatel41 , Anurag Dhingra and Snorre Kjesbu, @cisco described a different order. Developers got unlimited tokens first. Every employee got a personal agent next. Only after the habit formed did leadership ask teams to show value for their token spend, and now they are moving to per-employee token budgets. Controls came after fluency, and they came with a number attached. Most of the enterprises I work with started from the other end. They wrote the AI policy and stood up a review board before anyone outside the innovation team had used an agent for real work. Adoption stays thin, and governance ends up covering activity that never shows up. Isaac also points out that most organizations aim AI at productivity and efficiency, and growth, customer experience and AI-enabled products get neglected. I’d call that a culture problem too. Teams chase whatever leadership measures. If hours saved is the only AI metric on the scorecard, nobody will put their curiosity into a new customer workflow. If you are setting 2027 budgets now, put a per-employee token line next to the license line, and tie part of it to one customer-facing outcome. drive.starcio.com/2026/10/ai… via @star_cio