Generative AI is becoming an ever bigger part of our lives. Yet we continue to underinvest in the skills people need to use it well.
The opportunity we're missing
Look at how generative AI is actually being used today, and what stands out is how limited that use still is. Most people do not use it at all. Those who do tend to use it for one-off tasks: summarising a meeting, or turning a few notes into a polished email. Useful, certainly, but these are things that would have been done without AI too, just more slowly. The number of people using AI every day to do more complex work at scale is growing, but remains small. That is a missed opportunity.
What makes the current generation of AI models interesting is not simply that they can do existing work faster. They make it possible to do work that previously did not happen at all, because doing it at scale was simply impractical.
A manufacturer, for example, can use AI to search ten thousand technicians' service reports for recurring faults, reports that until now might only have been opened when a complaint was received. In the past, this kind of analysis was done on a sample basis, or not at all. Applications like these can be created by a small team of AI experts.
But the gains do not lie only in large, complex applications. They can also be found in everyday, simpler tasks that can now be carried out at a scale that was previously impossible.
Today's AI agents can take a request, a "prompt" in tech jargon, and independently perform multiple steps towards a final goal. A first draft of an email to a customer can now be written using the full context of what the CRM knows about that recipient, tailored to the specific situation and written in the sender's own style. The AI agent can also provide suggestions based on the customer's entire history. All that remains is to review the result, make a few adjustments if necessary, and send it. It is humanly impossible to conduct that level of research for every piece of correspondence.
And these capabilities are not confined to a handful of technical professions. Researching, comparing, summarising, analysing and writing are components of work across much of the economy. That makes AI skills relevant far beyond the people building AI systems themselves. Recent research on workplace skills similarly points to AI literacy, adaptability and broader cognitive skills becoming relevant across a wide range of occupations (International Labour Organization).
That is the leap: doing both new and familiar tasks faster, and with a level of attention that was previously impossible at scale.
But that scale also creates new risks. Anyone who asks AI to summarise ten thousand service reports without checking the resulting summary may have saved time, but they have also introduced a new risk. And anyone who gives a model access to a customer's complete history needs to understand why and how that data is being used.
That requires knowledge. You need to understand what you are delegating, how to verify what comes back, and when not to trust the answer. These skills build on each other, much like literacy does: over time, you learn a new way-of-working.
The grammar keeps changing
But there is an important difference. You can become fluent in English or Spanish and, broadly speaking, remain fluent.
AI is a language that keeps rewriting its own grammar. The systems change, their capabilities change, and the ways in which they fail change too. AI literacy therefore cannot be treated as a course you complete once. It requires continuous learning. This idea of AI literacy as an ongoing process rather than a fixed competence is increasingly reflected in the research, such as this 2026 systematic review of generative AI literacy across education and business in the International Journal of Educational Technology in Higher Education.
It is entirely possible that parts of the AI market are in a bubble. But that is a separate question from the long-term value of the technology itself. Even if expectations and valuations correct sharply, the ability to use these systems effectively will remain valuable.
AI literacy is becoming infrastructure
For employers, investing in upskilling their people in AI is a matter of self-interest.
But AI literacy is no longer just a corporate training issue. It is increasingly becoming an individual critical skill as well as a global public-policy theme.
The European Union is explicitly promoting AI literacy among workers and the wider population, while the OECD argues for expanding lifelong learning and making AI literacy accessible beyond specialists.
We take it for granted that everyone should learn to read and write, and that those who struggle with those skills should receive support. Within a few years, being able to instruct, assess and challenge an AI system may be just as fundamental to participating fully in working life.
If that is where we are heading, AI literacy cannot be left to employers and early adopters. We need to start investing in it across society now.
Alex van Gennep is co-founder and CEO of kju. He writes about how AI is actually landing inside companies, based on what leaders tell him. This piece was originally published on Alex's Substack on 26 August 2026.