The CTO’s AI Playbook Part 3: The Room in Which AI Decisions Are Made

​I want to tell you about a conversation I’ve had in various forms, many times over the past two years.

A CTO presents an AI strategy to the executive team. It’s thorough. The use cases are credible. The phasing is realistic. The risks are acknowledged. And then someone asks: “How long before we see ROI?” and the room stalls. Not because it’s a bad question, but because the people asking it and the person answering it are working from completely different mental models of what AI actually is and how it actually behaves.

This is what I call the leadership fluency gap. And in my view, it’s the most underrated barrier to enterprise AI success.

The data backs this up. According to the IBM CEO Study 2025, only 25% of AI initiatives have delivered their expected ROI and just 16% have scaled enterprise-wide. Those are sobering numbers for something receiving such enormous investment. And the Deloitte Tech Exec Survey 2025 found that 45% of tech leaders identify GenAI skills as the most urgently needed competency in their organizations with 70% planning to increase headcount specifically because of AI demands.

At Experis, we sit at the intersection of technology and talent, and what we see every day reinforces this. Our IT World of Work 2025 Outlook found that 76% of IT employers worldwide are struggling to find the skilled tech talent they need with AI now ranking as the second most in-demand skill globally, just behind cybersecurity. The Experis Tech Talent Outlook Q1 2026 reinforces it further: AI is no longer just a tech sector problem. It’s reshaping labour markets in finance, manufacturing, healthcare, retail everywhere. There is no safe corner of the economy that isn’t competing for the same small pool of people.

But here’s the thing the technical skills shortage, as real and serious as it is, isn’t the whole problem.

Technical skills get AI projects built. What gets them funded, resourced, sustained, and scaled is something different: leadership fluency.

When the CFO doesn’t understand why a foundational data investment has to come before the AI model, they push back on the budget. When the Chief Risk Officer doesn’t understand the difference between a deterministic application and a probabilistic one, they block deployment. When the CEO doesn’t understand why “the AI is ready but the organization isn’t” is a real and legitimate diagnosis, they lose patience at month three.

The CTO ends up spending half their time translating between what’s technically true and what the business expects at every single decision point. That’s exhausting. And it slows everything down.

I genuinely don’t think this is a failure of intelligence on anyone’s part. AI has moved fast faster than any normal continuing education program could keep up with. The vocabulary is specialized. The gap between what AI can do in a demo and what it reliably does in production is wide and genuinely difficult to calibrate. And the hype cycle has made it harder, not easier, for senior leaders to develop a grounded view.

But the impact is real. Programs get underfunded because their value is misunderstood. Timelines get compressed because the sequencing of foundational work isn’t appreciated. And when the results don’t match expectations which, given the expectation, is almost inevitable confidence drops across the whole AI agenda.

According to McKinsey’s State of AI 2025, leadership readiness remains one of the most significant barriers to AI maturity. Leaders, the research says, are “not steering fast enough.” I’d add: in many cases, they’re steering blind.

Here’s the counter-intuitive observation I’d add and it’s one that makes technically oriented people uncomfortable: organizations with the most technically elite AI teams often have some of the worst business outcomes from AI. Why? Because when the team is optimizing for model sophistication and the business is hoping for operational impact, you get elegant solutions to problems nobody prioritized. I’ve seen organizations build genuinely impressive AI systems that sat unused for months because the business teams who were supposed to adopt them had no idea why they should. Technical excellence without business fluency is a very expensive way to produce shelf ware.

So what does good actually look like?

The organizations making consistent progress tend to invest in AI literacy right across the leadership team not deep technical training, but focused education on how these systems work, what they can’t do, how to evaluate AI claims and how to ask the right questions. Two days well spent on this saves months of misaligned decisions.

They also build what I’d call cross-functional AI ownership governance groups that include legal, risk, finance, HR and operations alongside technology. AI decisions made only by the tech team tend to stall at the rollout stage. Decisions made by a team that represents the whole business tend to land faster.

And the best CTOs I know have invested in internal AI champions people inside each business function who understand enough to bridge the gap between technical teams and business teams. These often aren’t the most senior people. They’re the ones who get curious, learn fast, and speak both languages.

They’re invaluable. And they rarely show up on a traditional org chart.

AI transformation isn’t a technology project with leadership buy-in. It’s a leadership challenge that technology enables. That distinction matters because it changes who’s responsible for success.

A genuine question rather than a comfortable one: if your board had to sit an AI literacy assessment tomorrow, would you be confident in the results? And if the honest answer is no what’s the plan?

Next in the series: Part 4 Shadow AI: the risk your security team hasn’t budgeted for.