The Next Skills Gap Is Not What You Think It Is

Picture the scene. Your organisation has invested in AI tools. Copilot is rolled out. Access is there. A memo has gone round about responsible use. And somewhere in the building, right now, a pile of fifty pound notes is blowing out the door and into the street.

Not because the tools do not work. Because the gap between having access to AI and genuinely working with AI is larger than most organisations have acknowledged. And almost nobody is talking about what it actually takes to close it.

The moment organisations declare victory

Here is what I see constantly. An organisation invests in access. Microsoft Copilot rolls out. A few sessions happen. Someone sends a memo about responsible AI use. And then someone at the top says: we have done AI now.

They have not. They have bought the gym membership.

Most people at this point are using AI the way they use a search engine. Query in, answer out. They get something useful, they copy it, they move on. That is a perfectly natural starting point and it does produce real value. But it is nowhere near what is available, and the distance between where most people are and where they could be is significant.

What closes that distance is not more access. It is the willingness to push through a genuinely frustrating phase. The outputs are not right first time. Iteration feels slower than just doing it yourself. The thing does not seem to be working. Without support or preparation, most people interpret that discomfort as the tool failing rather than as a learning stage. So they step back. The tool becomes a novelty rather than a partner.

I have seen this play out so many times. I have also seen the moment it flips.

The Tony moment

We were running a programme at an automotive retailer. There was a finance director in the room, Tony, who had made it clear from the start that he did not want to be there. He thought this was a waste of his time. He had said as much to his colleagues.

By lunchtime he was practically dancing on tables. Not because we had told him something clever. Because we had sat him down, opened a laptop and shown him what AI could do with the actual problems sitting on his actual desk. The specific things that were eating his time and his energy. He had gone from sceptic to evangelist in half a day.

That does not happen through a generic training session. It happens when learning is specific enough to the person and the role that the penny drops in a real way. When someone stops thinking about AI and starts thinking with it.

The trust tax

There is a failure mode I want to name because it is expensive and it is happening everywhere.

Organisations put AI tools in. They make the investment. Six months later they are baffled because costs have gone up, not down. When you look closely at what has happened, the answer is almost always the same. The humans are quadruple-checking every output. They have no idea how the system works, what it is actually doing, or whether to trust what it produces. So they verify everything manually. Every time. The productivity gain has been entirely consumed by the trust tax.

This is what happens when you give people tools without giving them fluency. Not just skills. Fluency. The ability to understand what AI is doing well enough to know when to trust it and when to push back.

Our colleague Callum, who leads the psychology side of our work, puts it simply. If you give people prompts, it might work right now. In a few weeks or months it will change. What endures is how people think about it. Curiosity. Experimentation. The ability to work with AI rather than around it. That is what we are trying to build.

The next thing is already arriving

Here is what I want you to hold onto from this piece, because it is why this matters now and not in two years’ time.

The last few months have seen an exponential leap in what is called agentic AI. Not AI that waits for a prompt. AI that acts on goals. Systems that plan, execute, make decisions and complete multi-step tasks without someone holding their hand through every stage. We are building one ourselves, a synthetic chief of staff called Roxy, and the difference between that and a chatbot is not incremental. It is categorical.

Organisations that have not yet built genuine AI fluency in their teams are about to be asked to supervise systems they do not understand. To decide when to trust an autonomous process. To know what questions to ask when something goes wrong. To govern something they have never had to govern before.

That is a fundamentally different challenge from writing a good prompt. It is the next skills gap. And it has a name: agentic literacy.

What fluency actually looks like

We are seeing the difference in the organisations we work with. The ones navigating this well are not necessarily the ones with the biggest budgets or the most advanced technology. They are the ones that have treated AI adoption as a cultural project, not a software rollout.

They have built understanding from the board down. They have given their people learning that is specific to their roles and their actual problems, not generic awareness sessions. They have been honest about the frustration curve and supported people through it rather than expecting the discomfort to just resolve itself.

And crucially, they have not declared victory too early.

The work most organisations have done on AI literacy over the last two years is not wasted. It is the foundation. The question is whether you are building on it deliberately or hoping it will be enough.

A word on what SUSY is for

We built SUSY because we kept seeing the same gap. Organisations wanting their people to develop real AI fluency. Learning that was specific to role, to sector, to the actual work people do. Progressive enough to take people from basic confidence to genuine capability, not just an awareness session and a hope for the best.

The next version of that challenge is agentic literacy. Understanding what an AI agent is doing. Knowing when to direct it, when to correct it and when to override it. Knowing what your governance obligations are when an autonomous system surfaces something that requires a decision.

That is where we are going next. Because it is where the sector is going, whether it is ready or not.

Andrew Whyatt-Sames is Co-Founder of uptakeAI. uptakeAI builds AI literacy platforms and AI operating partners for organisations navigating sustained AI adoption.

To find out more about SUSY: susylearningai.co.uk