The CEO's Guide to AI: what I told a room full of CEOs in Chicago
Updated: 10 hours ago
I opened my keynote at the Cottingham & Butler CEO Summit in Chicago with a request that doesn't usually go down well in a room full of chief executives. I asked them to be willing to look foolish.
Think about the ten-year-old you who got hold of the family video recorder. Nobody read the manual. You pressed every button until it did what you wanted, and you couldn't break it. That's how humans learn new things fast. Somewhere between then and the corner office, most of us stop doing it. And you cannot build an AI capability passively, by watching, listening, and taking notes. So I gave everyone five minutes and a live tool, and we built things. Then we talked.
Here's the argument I made, in the order I made it.
A CEO has two jobs with AI, and most are only doing one
The first job is the one every board is asking about: apply AI to projects and get a return. The second is quieter, and it has to come first: build your organisation's AI capability in the first place.
Mid-market firms skip the second job. That's why the oft-quoted figure that 85 per cent of AI projects never deliver a return doesn't surprise me at all. The number has been kicking around since Gartner predicted it in 2018 (Gartner, 2018), and the recent evidence is no kinder: RAND puts the failure rate above 80 per cent (RAND, 2024), and MIT's 2025 study found 95 per cent of enterprise generative AI pilots delivering no measurable return (MIT NANDA, 2025). Nobody had an AI capability before 30 November 2022 (OpenAI, 2022). Not you, not your competitors, not the consultants selling you the roadmap. We all started from zero on the same day, and this hasn't crept up on anyone. There was one day, and it switched on. I've called it our Viking moment (see Our Viking moment), and it's the story I told on the TEDx stage (see Don't miss the boat twice). Expecting a return on a capability you haven't built yet isn't a technology problem. It's a sequencing problem.
You're the queen bee
I asked the room to bear with me on an analogy, and I'll ask you to do the same.
A beehive has a queen, who sets the direction. It has worker bees, who go out to the flowers they already know, bring back nectar, and make honey. Business as usual. Efficiency. Return on effort. And it has explorer bees, who fly off to find new flowers and come back to tell everyone where they are with a little waggle dance.
A hive that is one hundred per cent workers looks fantastic on the numbers. Accountants love it. Then the nectar runs out and the hive dies. A hive that is one hundred per cent explorers makes no honey. Evolution has already run this experiment. Depending on how much forage is about, somewhere between 5 and 35 per cent of a colony's foragers are scouts (Seeley, 1983); on stage I rounded that to 85 per cent workers and 15 per cent explorers.
Now look at your organisation. A time and motion study I quoted on stage puts most organisations at about 99 to 1, and three years of watching them closely tells me that's right. Not because people lack curiosity, but because targets, performance reviews, and bonuses all reward the worker bee. Nobody is measuring or incentivising how many explorers you have. And the field they'd be exploring is, right now, a garden of Eden nobody's looking at, because for thirty years there wasn't much out there worth finding.
Who sets the ratio? The CEO does. You're the queen bee. If you don't set it, nobody will, and business as usual will dominate.
What a capability is
Capability isn't a licence you buy. It's your best people, with the best tools, connected to your data (Dixon, 2024c). When you have those three, you're cooking.
I walked the room up the ladder. A model is the brain. A bot is a model with your data and a role. An agent is a bot in a harness: tools, search, your brand guidelines, sub-agents it can rustle up when it needs help, and, critically, your meeting recordings and transcripts. Then I showed them the harness working, live, on their own businesses. Competitor dashboards, target-client lists, and opportunity maps for delegates in the room, built from prompts that took me half a day to get right and now run in ten to fifteen minutes for anyone on my team.
My guess is that perhaps one organisation in a hundred is working like that today. The prompt is the product. Once you've built it, the whole team can run it.
You don't spec a bike. You ride one
Unclear purpose came third in the pre-event survey, behind lack of skills. I understand it, and I think it's backwards.
If you've only ever walked, and someone hands you a bicycle, you'll imagine bicycle uses within the twenty-mile radius you already know. You won't imagine the Tour de France. The way you learn what a bike can do is to ride it. So don't write the AI strategy first. Upskill your people, let them play for a month, and then ask them how they'd use it. Four months in, gather your best explorers and ask what you're going to do as a business. You'll discover the answer rather than trying to work it out in advance. In my experience it's a better answer than any you'd have written on day one.
The things I'd stop doing tomorrow
I gave the room a short list of don'ts, and they're worth repeating (the long version is in Stop Putting IT in Charge of AI):
Don't delegate the upskilling. Your fluency sets the ceiling for everyone else's.
Don't put IT in charge of the transformation. IT are looking for a spec. It isn't their job to invent your products and value propositions.
Don't wait for clean data.
Don't start with security and governance. The safest AI system is no AI system, in the same way that the safest way to cross the road is to stay at home.
Don't buy what a prompt could build, and don't outsource the whole thing to people who'll charge you for a spreadsheet.
Me, my team, my business
I closed with three levels, because an argument is easier to act on when it comes with a structure.
Me: lead the learning, set the ratio, and listen to the waggle dance. Two hours a week, personally, in the tools.
My team: amplify your explorers. Give them paid accounts, access to data, and permission to change things. Support the cautious, challenge the blockers, and reward capability when you see it. In any workforce of a hundred, around ten are already brilliant at this. You should be able to name them. If you can't let them fly, they'll fly somewhere else. Lock them in, because the whole world is about to want those people.
My business: build the capability, then apply it to projects, then look for the return. In that order. It's the sequence the whole of The AI Transformation Playbook is built around.
The homework was simple. Find your five most talented explorer bees. Go and look at their screens. Show them what you built. Then set the ratio.
Watch the keynote
The full hour is on YouTube, including the live demos, the delegates' own builds, and the moment I told a room of American CEOs they were the queen bee.
If you'd like to talk about where your organisation sits on the worker-to-explorer ratio, get in touch. I'd like to hear where you're up to.
References
Dixon, R.J., 2023. Don't miss the boat twice: my TEDx talk on the ChatGPT challenge. [online] robdixon.ai, 15 December. Available at: https://www.robdixon.ai/post/dont-miss-the-boat-twice-tedxwarrington [Accessed 3 September 2026].
Dixon, R.J., 2024. Our Viking moment: why the winning strategy changed on 30 November 2022. [online] robdixon.ai, 4 September. Available at: https://www.robdixon.ai/post/our-viking-moment-strategy-change-ai [Accessed 3 September 2026].
Dixon, R.J., 2024c. Defining AI Capability: A Holistic Framework for Organisational Readiness in an AI-Driven World. [online] LinkedIn. Available at: https://www.linkedin.com/pulse/defining-ai-capability-holistic-framework-readiness-ai-driven-dixon-omrle [Accessed 3 September 2026].
Dixon, R.J., 2025d. The AI Transformation Playbook: A Step-by-Step Guide to Becoming an AI-Enabled Organisation. Manchester: Dixon AI Press. Available at: https://www.amazon.co.uk/dp/1919400508.
Dixon, R.J., 2026. Rob Dixon | The CEO's Guide to AI | Cottingham & Butler Summit Chicago. [online video] Robert Dixon, YouTube, 11 June. Available at: https://www.youtube.com/watch?v=iS6sIvP0S8Y [Accessed 3 September 2026].
Dixon, R.J., 2026i. Stop Putting IT in Charge of AI (and 8 Other Mistakes Organisations Keep Making). [online] LinkedIn, 8 August. Available at: https://www.linkedin.com/pulse/stop-putting-charge-ai-8-other-mistakes-organisations-rob-dixon-ge0he [Accessed 3 September 2026].
Gartner, 2018. Gartner Says Nearly Half of CIOs Are Planning to Deploy Artificial Intelligence. [online] Gartner Newsroom, 13 February. Available at: https://www.gartner.com/en/newsroom/press-releases/2018-02-13-gartner-says-nearly-half-of-cios-are-planning-to-deploy-artificial-intelligence [Accessed 3 September 2026].
MIT NANDA, 2025. The GenAI Divide: State of AI in Business 2025. [online] MIT NANDA, July. Available at: https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf [Accessed 3 September 2026].
OpenAI, 2022. Introducing ChatGPT. [online] OpenAI, 30 November. Available at: https://openai.com/index/chatgpt/ [Accessed 3 September 2026].
Ryseff, J., De Bruhl, B.F. and Newberry, S.J., 2024. The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed. [online] RAND Corporation, 13 August. Available at: https://www.rand.org/pubs/research_reports/RRA2680-1.html [Accessed 3 September 2026].
Seeley, T.D., 1983. Division of labor between scouts and recruits in honeybee foraging. Behavioral Ecology and Sociobiology, 12(3), pp.253–259. Available at: https://link.springer.com/article/10.1007/BF00290778.
Rob Dixon is the founder and CEO of Dixon AI and speaks regularly on AI, strategy, change, and innovation. The Cottingham & Butler CEO Summit took place in Chicago in 2026.
Comments