Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Most leaders think they’re AI fluent

Their usage says otherwise.

[Photo: Getty Images]

Profile photo of Daniele Grassi

CEO · General Assembly

Ask a senior leader how fluent they are in artificial intelligence and you’ll usually get a confident answer. Ask what they did with it this week, and the answer gets a lot thinner.

The distance between how leaders rate themselves and how they use the technology was one of the clearest findings in new research from General Assembly, Ezra, and LHH brands. It focused on how business leaders are approaching AI at work. Drawing on more than 500 directors, vice presidents, and C-suite leaders across the U.S. and U.K., the survey found that leaders’ AI confidence is real and rising, but their practice behind it hasn’t caught up.

THE GAP BETWEEN CONFIDENCE AND USAGE

On paper, leaders look highly AI fluent: 83% of leaders now say they have the knowledge to make AI-related vendor decisions inside their function, up from 66% in 2024; 68% say they understand how to deploy AI without putting company data at risk; and 71% believe they can accurately judge how competent their own employees are with the technology. By their own scorecard, this is a generation of executives who feel ready to buy AI, govern AI, and evaluate everyone beneath them on it.

Usage tells another story. We then asked a simpler question: What did you actually use AI for this week? The results showed a lack of AI fluency: 69% said searching for information; 68% said summarizing documents; and 58% said drafting emails.

For the strategic work you’d expect competence to unlock—scenario planning, financial modeling, and organizational design—the numbers fell below 30% across the board.

The same leaders making seven-figure procurement calls and signing off on governance frameworks are, by their own account, mostly using AI for tasks a search engine handled a decade ago. To be clear, there’s nothing wrong with using AI that way. Faster summaries and cleaner first drafts are real gains, and not every task needs reinvention. The trouble is that using a tool for its most basic functions tells you very little about whether you understand it well enough to make serious decisions about it.

A leader who only ever asks AI to tidy an email has no real feel for where it breaks, where it quietly invents things, where the data risk lives, or what “good” looks like when a team uses it for something that matters. The kind of understanding required for vendor selection and governance decisions only develops through deeper, hands-on use.

WHY THE GAP IS EASY TO MISS

This is the uncomfortable part of the finding: The confidence isn’t necessarily false, but it’s running ahead of the experience that’s supposed to earn it. Leaders are rating their fluency based on having access to the tools and using them comfortably, not on having pushed them anywhere demanding. And confidence built on shallow use is exactly the kind that doesn’t know what it doesn’t know.

So, what separates leaders whose confidence is backed by capability?

The data points to one answer more clearly than any other. Leaders whose self-assessment and real practice lined up had been through structured, leadership-specific AI training. They were far more likely to say they could deploy AI without compromising company data (88% versus 68% overall). They were also more likely to have restructured their teams around the technology (68% versus 52%). And they were nearly three times as likely to have built a working tool themselves (49% versus 19% of leaders without training).

That last number is one I keep coming back to. Building something, even something small, is the moment abstract familiarity turns into real understanding. A leader who has shipped even a rough working tool knows from experience where AI helps, where it struggles, and what it costs to run. It’s a different kind of knowledge than being able to describe what the technology does.

Training is what moves people across that line. Leaders who’ve crossed it stop guessing about the technology and start making decisions grounded in how AI behaves in operation.

The kind of training that does this looks nothing like the generic course catalog most companies hand out. Long, self-paced video modules and abstract use cases can lift a completion rate while leaving the underlying gap exactly where they found it. What closes the gap is training built for the decisions leaders make every day, taught in their real tools and constraints, with enough hands-on building that fluency stops being a feeling and becomes something demonstrated.

A BETTER QUESTION THAN “HOW CONFIDENT AM I?”

None of this means leaders should be embarrassed about where they are. The honest move is just to stop treating comfort with a tool as command of it. The questions worth asking are quieter and more useful: When did I last use AI for something that genuinely tested it? Have I ever built anything with it, or only asked it for favors? If a vendor walked in tomorrow, would my judgment rest on experience or on a sense that I ought to know this by now?

Leaders who will come out ahead aren’t the most self-assured ones in the room. They’re the ones whose confidence has something underneath it, who’ve done the work to make their fluency real rather than assumed. For some leaders that gap between believing and doing is starting to close.

For plenty of other leaders, AI competency is still being papered over with certainty they haven’t earned. If you’re unsure where you stand, ask yourself the questions above, and be honest. If the gap is there, work to fix it, or live with the fallout.

Daniele Grassi is the president and chief executive officer of General Assembly.