Every chief information officer I talk to describes the same thing: a mandate to own AI, and no one above them deciding what it's for. A panel at Creatio No-Code Days Florida 2026 described, almost without naming it, what fixes that.

Ask a chief information officer what their biggest AI problem is, and the answer is rarely a model or a vendor. The real answer is closer to this: someone above them said go own AI, and never said what for. No requirements arrive. No one above them ranks what matters first. The mandate shows up. The map does not. That gap was the real subject of the AI org-design panel at Creatio No-Code Days Florida 2026, held at The Ritz-Carlton Orlando, Grande Lakes, even though no one on stage used those words.

The panel had range. Janet George, executive vice president of artificial intelligence at Mastercard. Angela Stewart, vice president of corporate systems at Nasdaq. Chris Pallavicini, a chief information officer at Aon. Tomy Han, partner at Volition Capital, an investor in Creatio. And Dr. Ja-Naé Duane, bestselling author and AI researcher tied to MIT and Brown University.

A personal note. I have known Ja-Naé Duane and Angela Stewart for years, and I have heard Stewart talk about this exact problem at other conferences: helping the business solve workflow bottlenecks, with AI or without it, before anyone argues about the technology. Hearing her make the same point again here told me it is not a stage line. It is how she works.

I ran an analytics platform inside a much larger company years ago, and every business unit wanted something different from it, on its own timeline, with its own idea of what mattered most. No one above the platform team ever ranked those requests. The platform team ended up doing the ranking by default, and took the blame when the ranking was wrong. That is the same shape of problem most CIOs describe to me now about AI, just with higher stakes.

George's answer at Mastercard splits the job in a way that fixes exactly this. She runs a hub-and-spoke model. The hub owns the shared platforms, the clean data, and the rules: who can see what, what gets tracked, how bias gets checked. The business units own the use cases, because they sit closest to the customer and closest to the cost of getting it wrong. The hub does not rank the business's priorities for them. It makes sure that whatever the business picks can be built safely and reused.

That single move, governance with the platform team and prioritization with the business, is the part most organizations skip.

Stewart described how this plays out at Nasdaq. She runs two speeds at once: teams improving daily work step by step, with the rules team close by, and what she calls moonshot factories, built to skip small steps and aim for big leaps. The connector between the two is the sandbox, a safe play space cut off from live systems.

The sandbox flips the usual order. Instead of the tech team pushing tools at the business and guessing what matters, the business comes to Stewart and says, I built this over the weekend, imagine what we could do with real help. That is a priority list, generated by the people who feel the bottleneck, instead of guessed at by the people running the platform.

Han added the investor's version of the same point. Across Volition's companies, the sandbox works like a talent test. Open it to everyone, and leadership is often surprised by who steps up. Some companies are now rebuilding teams around the people who rise, not the old chart. Titles hide talent. Sandboxes reveal it.

Pallavicini described Aon's training program: more than 60,000 colleagues, top floor to service desk, covering not just what AI is but how the models actually work, with a steady message that AI adds to the role rather than replacing it. Olkowski's reaction was immediate. What he usually hears from other companies, he said, is the opposite: leadership telling teams to just use AI, with nothing behind it. You could hear the contrast land in the room. That program has a number, an owner, and a story anyone in the room could repeat.

George's point was about what happens after the map and the training are both in place. Once agents go live, she said, one bad act by one agent can break trust in the whole program. Trust is not something you grant. It is a design choice, built into the platform the agents run on, from day one, not patched in after the first mess. What the platform watches, what it logs, what it refuses to let an agent do, what happens in the first minute after an agent goes wrong: that is where trust actually lives or dies.

Creatio's own AI Studio demo earlier in the event put prompts, model choice, limits, and testing in one safe space, which is at least a step toward giving that layer a name. How deep those controls go is the question I will keep asking in briefings, and it follows the thread from my earlier post on Creatio's agentic platform.

Duane's closing point connects here too, though no one drew the line on stage. Olkowski checked the clock right before he handed her the floor: three minutes and ten seconds left, he said, for the part of the panel about the human cost of moving fast. Duane took the time anyway. Speed without care burns people out, she said. When leaders push faster without rest stops or review points, the bill comes due as burnout. It comes due another way as well. Review points are exactly where a bad agent gets caught before it goes live. Skip them for speed, and George's single bad episode stops being a hypothetical.

The panel had three minutes for that point. Most companies give it less.

Her advice: make pace itself a design choice, and respect the grief that comes with it. People who spent months going one way feel the pivot as a loss. Pretend otherwise, and the pushback you feared shows up anyway.

The closing round asked each panelist for one practical word. Communication. Experimentation. Foundations. And the one that stuck with me: unlearn. The room broke into applause before anyone could ask a follow-up.