A Gartner analyst published a guest post this week arguing that CIOs need to fix a problem that sounds simple but isn't: most executives don't understand AI well enough to lead it. The data behind that claim is worth sitting with, because it shows this is not a training gap. It is a trust gap at the very top of companies.
In a Gartner survey of more than 450 CEOs, less than half of CIOs were rated "AI-savvy" by their own bosses. The chief information security officer and chief data officer scored no better. A separate Gartner figure cited in the guest post puts the number even lower: just one in five executives sees their fellow executives as genuinely capable with AI.
That gap has a price tag. A widely cited MIT study found that the overwhelming majority of company AI pilots produce no measurable financial return, and the researchers pointed to poor planning and mismatched priorities as the cause, not weak technology. Put those two facts together and the picture is clear. Companies are approving AI spending decided by people who, by their own colleagues' admission, do not fully understand what they are approving.
This is why framing AI literacy as a personal skill, like knowing how to use a chatbot, misses the point. The real skill executives need is judgment: the ability to sit across the table from a vendor, hear a bold promise, and know which parts are realistic and which parts are sales talk. It also means understanding that AI is only as good as the company's own data, so a leader who skips that question will keep signing off on projects that quietly fail months later.
Gartner puts a number on the upside too, predicting that companies who invest in this kind of executive training will post noticeably stronger financial results within a couple of years than those who don't. That prediction lines up with what more executives are already doing. Survey data shows CEOs and chief financial officers have sharply increased their personal involvement in AI decisions over the past year, pulling that responsibility away from IT departments and into the boardroom.
This matters well beyond tech firms. A steel manufacturer, an insurance consultancy, or a retail group does not need a leadership team that can code. It needs a leadership team that can tell a genuine capability from a demo reel, and that can ask a vendor the one uncomfortable question that saves a year of wasted budget.
Putting this entirely on the CIO's desk is a mistake many companies are still making. Literacy about AI is becoming a basic leadership skill, closer to reading a balance sheet than to running a help desk. Boards that keep treating it as an IT problem will keep funding pilots that go nowhere, while the companies that treat it as a leadership problem pull ahead.