The government made a welcome announcement this week. A new AI and Automation Practitioner apprenticeship has launched, backed by a commitment to upskill 10 million workers with AI skills by 2030, with £187 million earmarked for digital and AI education. The aim, as Ministers put it, is to ensure the workforce can “adopt AI and make the UK a leading AI nation.”
It is hard to argue with the ambition. The numbers tell an uncomfortable story. We know that 97% of UK organisations report at least one AI skills gap, and 28% say those gaps are already impacting their ability to achieve business goals. Fewer than 20% of UK workers feel confident using AI tools in their daily role. Meanwhile, 70% of UK workers worry about AI’s economic impact, but only 39% believe their own jobs are at risk. This psychological disconnect is what researchers are calling the “AI perception gap”.
So yes, we need more training, but there is a structural problem with how we are approaching it.
Every apprenticeship, every upskilling programme, every lunch-and-learn session is aimed at the workforce. At the people doing the work. At the employees who will, rightly, need to adapt to AI-augmented roles. But almost none of it is aimed at the people deciding which AI to deploy, how fast, and with what governance in place.
I am talking about boards and senior leadership teams.
Here is what I see repeatedly in my work with executives and directors. AI tools are being procured, piloted and rolled out by technology functions while boards receive a brief update at the quarterly meeting. Strategy is being outsourced to whoever has the loudest internal voice on AI, often someone with genuine enthusiasm but no particular mandate to think about risk, ethics or accountability. Meanwhile the people at the top are quietly relieved that it is someone else’s problem, at least for now.
Here’s a telling gap: 23% of CEOs say their staff are already handing whole tasks to AI but only 8% of employees report doing this. Either CEOs are overestimating AI adoption in their organisations or they are not close enough to what is actually happening on the ground. Either way, it points to a leadership visibility problem.
This is not a criticism of individual leaders. AI is moving fast, the vocabulary is unfamiliar, and nobody wants to look foolish in the boardroom. But the effect is that governance decisions (about data use, about accountability, about the employment consequences of automation) are being made at the wrong level of the organisation, or not made at all.
The new apprenticeship will help practitioners understand AI tools, avoid bias and comply with regulatory requirements. All of that is valuable. But regulatory compliance is not a practitioner-level responsibility. It belongs to leadership and leaders who cannot distinguish between a legitimate AI use case and a governance risk are not equipped to discharge that responsibility, however many practitioners they have trained beneath them.
Organisations that invest in structured, organisation-wide AI literacy programmes are nearly twice as likely to report significant AI return on investment compared to those without such programmes. The operative word is organisation-wide. Not workforce-wide. Not entry-level-and-upwards. Organisation-wide, which means it has to include, and ideally start with, the people setting direction.
The UK government’s AI Skills Boost initiative is a genuine step forward. But the training trap is this: we keep pointing the investment at people who will use AI, while leaving the people who govern its use without the knowledge they need to do so responsibly.
If you are a senior leader reading this and thinking “that is probably fine, we have good people on it”, that response, well-intentioned as it is, is exactly what I mean. Good people cannot compensate for absent governance. And absent governance at board level is a risk that no amount of apprenticeship training will fix.
AI literacy is not a workforce problem. It is a leadership imperative. The sooner our skills strategy reflects that, the better.
