Watch Out For AI Ed

Like it or not, AI literacy programs are everywhere. American colleges have stood up roughly 1,000 AI programs, up from five schools with an AI major in 2021. The EU now obliges every provider and deployer of AI systems to ensure a “sufficient level of AI literacy” among their staff. The White House has collected pledges from more than 60 organizations to fund AI education for American youth. But what, precisely, is being taught?

The AI Literacy Trap

CEOs need to cast a skeptical eye on the proliferation of these so-called AI literacy programs and draw the line between what their executive team genuinely needs to know versus what falls under the purview of their CTO or technical staff. Knowledge of the differences in technical design between the Nvidia GeForce RTX 5090 and the Nvidia Vera Rubin chip is not necessary for effective use of AI.

Consider prior technological advances that revolutionized business but which CEOs did not need to master:

• No executive teams needed to be technically telephonically literate in circuit switching before placing a phone call.

• Millions of executives board aircraft every day without the faintest idea how a turbofan is bolted to a wing or the aeronautics of lift.

• Most of crypto’s largest investors would be hard-pressed to explain the mathematics behind blockchain development.

• Most industrial executives, let alone food titans, would find it challenging to explain the intricacies of energy transfer with chemical fluids in refrigeration.

Prior generations of executive teams knew where to draw the line between technical, deep domain knowledge and more general practical lessons for executives.

Then, there is the awkward matter of the curriculum itself. Not all programs live up to their promises. A systematic review published in npj Science of Learning found 16 competing AI literacy scales that agree on almost nothing beyond three headings so broad—technical understand ing, societal impact, ethics—they could describe a sophomore seminar on any subject invented since 1900. A 2026 article in Learning, Media and Technology is titled, without irony, “The (im)possibility of AI literacy.” The European cognitive scientist Carl Hendrick puts it best: The term has achieved “near-universal acceptance despite almost no agreement about what it actually means.”

Where the content does get specific, it is often outpaced by events. The skill at the center of nearly every corporate AI curriculum, prompt engineering, was a $200,000 job title in 2023 and a punchline by 2025, obsoleted by models that no longer need coaxing. Now, anyone using AI employs prompt engineering by default, with no specialized training needed. A discipline whose ground is shifting faster than anyone can track is not a body of knowledge that lends itself well to curricular development at this stage.

Issues with Instruction

Worse, the training may be actively harmful. In the Harvard–BCG field experiment on 758 consultants, AI was a triumph on tasks inside its competence: 25 percent faster, dramatically better output. But the model got one business problem wrong. Consultants working with no AI at all answered correctly 84.5 percent of the time. Those given the AI support fell to 70 percent. And those given the AI plus a prompt engineering overview—the literate ones, the trained ones—finished dead last at 60 percent. Instruction did not sharpen their skepticism; instead, it deepened their deference to sometimes wrong information, with curriculum-driven rigid, formulaic instructions overruling basic common sense and false mathematical precision regrettably triumphing over the flexibility to adapt to rapid changes.

So stop asking whether your workforce is AI literate. The real question is whether your workforce is integrating AI into workflows, questioning and confirming findings, mindful of groupthink pathologies in LLM training and alert to dangers of irresponsible uses of any technology. The curriculum being sold by schools and commercial conferences preying on business’ AI anxieties and eager to brand themselves as AI resources is at risk of becoming the snake oil of cyberspace.