A couple of days ago, I was sitting on the floor of my daughter’s room, with my coffee, watching her face-plant into the yoga mat for the fourteenth time in a row. She’s three months old. Learning to crawl. Zero frustration. Belly down, arms out, and back into the mat. Again.
And I had this thought I couldn’t shake: my three-month-old daughter learns better than most Fortune 500 companies.
And I don’t mean that as a cute metaphor. I mean it structurally. And as a behavioral economist by education — I studied economics at Bocconi — I can’t help thinking that the way a baby learns to crawl is the best way to think about behavioral change. And since cultural change is nothing more than collective behavioral change, crawling contains every single principle an organization needs to adopt an AI-first culture.
You know what nobody tells you? Babies don’t learn to crawl because someone teaches them. You create three conditions, and the crawling emerges on its own.
And that word — conditions — is the one that 95% of companies, according to a recent MIT Media Lab report, are getting wrong right now in their AI transformation journey.
And today we’re going to talk about why AI-first is a crawling problem, not a technology problem.
Let’s start with the baby. In developmental science, crawling is not a motor skill. It’s an emergent behavior. A motor skill is something you train. An emergent behavior appears when the right conditions exist.
What are the conditions?
There are 3:
First: tummy time. You put the baby belly down. It’s uncomfortable. But that discomfort activates exactly the muscles she needs.
Second: you let her fail. She falls fourteen times. And nobody files an incident report. The cost of failure is zero.
Third: you place a toy. Just out of reach. So she has a reason to try.
Now. Pause here for a second.
A study by Ying Bao and colleagues, presented at the Hawaii International Conference on System Sciences, investigated what makes people trust AI during collaboration. And what they found maps almost perfectly onto the baby. Two dimensions: cognitive perception — is this tool easy to use? Is the coordination cost low? And emotional perception — do I feel comfortable? Do I enjoy it?
Tummy time reduces cognitive complexity. Letting her fail reduces coordination cost. And the toy is emotional comfort and enjoyment.
Three human conditions. And most companies have zero of the three.
And what have the companies winning with AI actually done? They’ve redesigned four things. Not their tech stack. Four dimensions of how the organization thinks, builds, serves, and grows. I call them the four floors — because you can’t crawl on a broken floor. And I summed it all up in my book “Between You and AI.”
Floor one: Data. From reactive to predictive.
Mayo Clinic is number one in CB Insights’ AI Readiness Index. They don’t treat disease. They prevent it. Your dashboards: are they a rearview mirror, or a windshield?
Floor two: Design. From automation to augmentation.
Nubank didn’t replace their human agents with chatbots. They built an AI-powered panel that makes those humans better — real-time context, sentiment analysis, suggested next actions. That’s not automation. That’s augmentation. And it’s a design choice, not a technology choice.
Floor three: Value. From transactions to experiences.
John Deere. For 187 years, a tractor company. Today, they sell intelligence to the farmer. The tractor is the same. The value proposition is completely different. Stock up over 500%.
Floor four: People. From hard skills to soft skills.
Scott Aaronson said something that should keep every professional awake at night: for any task with an objective measure of success, it’s just a matter of time before AI outperforms the best humans. What survives? Judgment. Empathy. Ethical reasoning. The ability to sit in ambiguity and still decide.
Four floors. Data, Design, Value, People. None of them are technology problems. Every single one is a leadership problem, a design problem, a trust problem.
And here’s where this article made me lose sleep.
Deloitte’s TrustID Workforce Index found that senior leaders are 40% more likely to use unapproved AI tools than their own staff. In the tech sector, shadow AI usage hits 58%. The people who are supposed to be leading the transformation are the ones bypassing the system the most.
But it gets worse. Employee trust in company-provided generative AI fell 31% in just three months in 2025. Trust in agentic AI — systems that act autonomously — dropped 89%.
And Amelia Dunlop, CXO at Deloitte Digital, stated that workers see their employer as two times less empathetic after AI is introduced into the workplace.
We’re deploying AI faster than ever. And people trust it less than ever.
So, what’s the CTO’s role in all this? It’s four. One for each floor.
Data — Architect of the Foundation.
You decide whether your organization’s data is AI-ready or AI-hostile. Clean, centralized, governed, real-time. Boring work. Not sexy. But it’s the difference between scaling AI and scaling pilot projects.
Design — Ecosystem Designer.
Your job isn’t to build every tool. It’s to build the ecosystem in which tools get built, deployed, and governed. Approved model stack. Tiered risk classification. Vibe-coding governance. How many of you have a published policy on what marketing can build with AI? Exactly. That’s the gap.
Value — Strategic Partner.
This one is uncomfortable. It means leaving the server room and sitting in the boardroom. The brutal truth: the CTO who can draw a line from a model deployment to a revenue number keeps the budget. The one who can’t becomes a cost center. And cost centers get cut.
People — Champion of Human Readiness.
And here’s where the baby comes back. Remember the three conditions? Tummy time, permission to fail, and a toy. That’s your job. Not HR’s. Yours. Because you’re the one who understands what AI can do, what it can’t, and what it needs from humans. McKinsey found that companies where senior leaders actively role-model AI use are three times more likely to be high performers. Three times. Because of behavior, not technology.
Four floors. Four roles. Architect. Designer. Partner. Champion. And in none of them is the CTO the person who builds the AI. In every one of them, the CTO is the person who builds the conditions for AI to work.
Conditions. Not code.
Just like a parent building the conditions for a baby to crawl.
AI-first is not a technology strategy. It’s not about models. It’s not about compute. It’s not even about data, though data matters enormously.
AI-first is a crawling problem.
It requires discomfort — redesigning workflows, not bolting AI on top. It requires permission to fail. And it requires a reason to move.
And above all, it requires trust. Not the kind you declare in an email. The kind you earn through decisions that cost you.
My daughter will learn to crawl. She doesn’t need a strategy deck. She needs three conditions and a parent who creates them.
Your organization is no different. The algorithm can do a lot of things. But it can’t crawl. Only your people can. And they’re waiting to see whether you’ll create the conditions — or just send another email about AI.
That’s the question. What will you build? A program, or a yoga mat?
Check out my book “Between You and AI,” published by Wiley, and you’ll find everything about me at andreaiorio.com.
