Blog
Skills3 min read

Almost nobody using AI thinks they are good at it

The majority of people using AI at work do not believe they have the skills for it, and more than seven in ten want to learn. The constraint is not willingness and it is not access.

The upper galleries of a large historic library, shelves rising several storeys.
SkillsPhotograph via Unsplash

There is a story about AI adoption that goes: people are resistant, they need convincing, roll it out slowly and manage the change. It is a comfortable story because it makes the problem cultural, and cultural problems justify long programmes of work.

The data does not support it. What the data shows is a workforce that is already using these tools, does not think it is any good at them, and is asking to be taught.

The finding, in one sentence

70%
of workers who use AI said they were enthusiastic to learn more about it, even though the majority did not consider themselves to have specialised AI skills.OECD survey of 5,334 workers · OECD, The impact of AI on the workplace

Both halves of that matter and they are usually reported apart. Most people using AI at work do not think they are skilled at it. More than seven in ten of them want to get better. That is not resistance. That is an unmet training demand sitting inside your own building.

Nobody is refusing. They are waiting to be shown, and nobody is showing them.

Who does report having the skills

The survey broke self-reported AI skills down by group, and the pattern is the one you would guess, which is itself the point: fluency is concentrated exactly where a business already concentrates its advantages.

  • Workers with a university degree: 57% in finance and 59% in manufacturing say they have specialised AI skills.
  • Managers: 58% in both sectors.
  • Workers born abroad: 65% in finance and 49% in manufacturing, the highest single figure in the set.
  • Workers in finance aged 16 to 24: 55%.

A business that does nothing will end up with AI capability that maps onto seniority and education, which is to say onto the people who already had the most leverage. That is not a neutral outcome. It is a widening of an existing gap, produced by inaction rather than by decision.

The same gap, one level up

The generational split in the general population is even wider, and it runs straight through the middle of most workforces.

66%
of US adults aged 18 to 29 use AI chatbots, against 23% of those aged 65 and over.Surveyed 17 to 23 February 2026, 5,119 US adults · Pew Research Center

Forty-three points. Any organisation with a normal age distribution contains both of those populations, using different tools, at different frequencies, with different assumptions about what is normal. Treating that as one audience with one rollout is the most common mistake in this whole area.

What employers say they actually need

The counter-intuitive part, and the one worth taking to a board: employers who have adopted AI report that it increased the importance of specialised AI skills, and that it increased the importance of ordinary human skills and highly educated workers more.

That is the opposite of the replacement story. The organisations furthest into this are not reporting that they need fewer capable people. They are reporting that capable people matter more, because a system that produces a plausible answer in three seconds moves all of the value to whoever can tell whether the answer is right.

What to do with this

  • Stop budgeting for change management and start budgeting for training. The resistance you have planned for does not appear in the data; the skills gap does.
  • Train the people who did not go to university and are not managers first. That is where the gap is, and it is where the cheapest gains are.
  • Assume two populations, not one. A rollout designed for the 66% will lose the people who most need it.
  • Measure fluency, not access. Access is close to universal and it predicts nothing. Whether somebody can tell a good answer from a confident one is the whole game.

Sources

02

New writing, when there is some

A short note when something is published: what it is about and a link. No more than a couple a month, and nothing else.

Start with
the diagnostic.

A scoping conversation costs nothing and ends with a straight answer about whether there is enough here to be worth doing. If there is not, we will tell you.

Book a scoping callinfo@skaiyon.com

Questions. Asked before every engagement.

Your data is yours. You can export it at any time, and it is exported to you in a documented format before any engagement closes. The software itself is licensed: we build it around your business, host it, and run it, and you pay monthly for that. If you'd rather own it outright, that's possible. It's a different kind of engagement and it's priced accordingly. Licensing keeps maintenance our problem rather than yours, which is why it is the default.

Two weeks' notice, either side. Your data is yours. You can export it at any time, and it is exported to you in a documented format before any engagement closes. The system stops running. If you'd rather keep it running, the ownership option is available at that point as well. Ending the retainer doesn't force you to lose what was built.

Typically four to eight weeks from signing to a working system, depending on how many tools it connects to and what state the data is in. A scoping conversation and a written plan come first, so the timeline is agreed before anything is committed to.

Access to the tools and data the system will work with, one person who can make decisions, and a few hours in the first two weeks while we map how work actually moves through your business. After that, very little. The point of the engagement is that it runs without your attention.

Usually. Most business software exposes an interface we can build against, and where one doesn't there's normally a way around it. Which connections are viable is settled in the scoping conversation, so you find out before committing rather than after.

Hosting region for a system we build for you is decided at the start of the engagement, not inherited from a default. If data residency is a hard requirement, raise it in the first conversation and we will tell you plainly whether we can meet it. This website is separate and its processing is already fixed: enquiries, quote requests, bookings and chat are handled by Supabase in Tokyo, Resend in Tokyo, Cloudflare, cal.com and Moonshot AI, which the privacy policy names individually along with where each one processes. Nothing from a client system is ever sent to the chat assistant.

That's the normal starting point, and mapping them is part of the work. Automating a process nobody has examined just makes the confusion faster, so we don't start there. The first phase establishes how things actually happen, as opposed to how they're supposed to.