Count the column inches devoted to AI literacy and you will run out of fingers quickly. Barely a week passes without a think-tank report, a ministerial speech, or a LinkedIn thought-leader declaring it the defining skill of our age. The World Economic Forum says 44 per cent of workers’ core competencies will be disrupted by 2027. IBM reckons 35 per cent of the global workforce needs reskilling in the next three years, up from just six per cent in 2021. IDC, a research firm, warns that skills gaps could cost the global economy 5.5 trillion US dollars by 2026. The alarm bells have been ringing so long and so loudly that nobody has stopped to ask what, precisely, they are ringing about.
The core problem is that if you ask ten people what AI literacy means you will likely get ten very different answers. UNESCO describes it as the knowledge and skills needed to understand how AI systems function and their social, ethical, and economic impacts. The European Commission and the Organisation for Economic Development and Co-operation (OECD) jointly produced a framework breaking it into three broad domains: interacting with AI, creating with AI, and managing AI’s actions.
The US Department of Labor published its own framework in February 2026, carefully noting that “indicating a need for AI literacy is not enough on its own”. Employers need to specify the “specific AI skills and depth of knowledge appropriate for each role”. A candid admission, buried in bureaucratic prose, that the term as commonly used tells you almost nothing.
The result is, as one analyst at ETS, an educational testing organisation, put it bluntly, an “alphabet soup” of competing frameworks. All great, all slightly different, none of them interoperable. Nobody can measure whether a workforce is AI literate when nobody has agreed what that means. Skills cannot be verified. Credentials carry little weight. Training programmes proliferate, mostly selling broadly the same thing with different logos.
AI comfort
What employers actually want, when pressed, turns out to be more modest than the rhetoric suggests. Most workers do not need to understand transformer architectures or fine-tune a language model. What they need is to know how to give a useful prompt, evaluate what comes back, and spot when a machine has confidently produced something wrong. A 2026 survey by Resume Builder found that 72 per cent of hiring managers now factor AI comfort into decisions, up from 48 per cent in 2024. That is a significant shift in two years. But ‘comfort’ is doing a lot of lifting in that sentence. It is little more than the AI equivalent of knowing how to use a search engine.
Surprisingly, perhaps, only 17 per cent of employees use AI tools frequently in their jobs today, despite 42 per cent expecting their roles to change significantly because of AI within the next year. That gap, between anticipated disruption and current adoption, says more about confusion than complacency. Workers are being told to prepare for a transformation they cannot quite see yet, using skills nobody has cleanly defined, through training programmes their employers often expect them to seek out themselves. Forty-two per cent of employees say exactly that.
The sceptics’ premium
Employers are making things worse. Even as AI spending surges, training budgets are being cut. LinkedIn’s 2026 Workplace Learning Report found only 26 per cent of organisations now offer formal AI upskilling programmes, down from 35 per cent the year before. Companies are buying more AI and teaching less about how to use it. The logic, presumably, is that the tools are intuitive enough to pick up on the job. They are not.
The irony is that the most valuable component of AI literacy is the one that no EdTech vendor quite knows how to sell: scepticism. The ability to look at a plausible-sounding AI output and think, Actually, is that right? Teams that use AI without understanding it, as one consultancy put it, “often work harder, not smarter—spending hours fixing AI-generated errors, second-guessing recommendations, or avoiding useful capabilities entirely”.
PwC’s 2025 AI Jobs Barometer found that AI-exposed roles are evolving 66 per cent faster than comparable jobs, and command a wage premium of 56 per cent. Those gains do not go to workers who have completed an online module and clicked the certificate button. They go to people who understand enough about what AI can and cannot do to make better decisions with it. That is a skill. It just has not yet found a name clear enough to put on a curriculum.
Until it does, AI literacy will remain what it largely is today: a useful anxiety, a thriving industry, and an answer in search of a question.
Photo: Dreamstime.
