The AI apprenticeship is the right move. Now make sure it trains for the right thing

The Level 4 AI and Automation Practitioner apprenticeship launched by Skills England has arrived at a difficult moment for the FE sector. The management apprenticeship suite from Team Leader to Operations Manager and Chartered Manager has all been withdrawn from funding, just as the Department for Education reported a £6 billion drop in employer training spend between 2022 and 2024. At the same time, the government’s wider AI Skills Boost commitment, which aims to upskill 10 million workers by 2030, has reframed AI capability as a workforce-wide requirement rather than a specialist one.

Into that landscape, Skills England has put a single, sector-agnostic, employer-designed standard open to organisations of every size. I think it’s the right move. The question is not whether this apprenticeship is needed. It’s what gets trained for once it meets the open market.

What the hiring market has already learned

I’ve spent the last year looking at how employers are incorporating AI into hiring – what levels of AI fluency they expect, how they define it, and how often they get it wrong. The data is useful intelligence for anyone now designing this apprenticeship at scale, because the labour market has a head start on the FE sector and the early signs are not encouraging.

In a survey of nearly 2,000 senior hiring leaders across the UK and US, 95% listed AI competency as a stated hiring requirement. 71% said they had formally defined what AI fluency means for their teams. A further 53% said they now prefer a candidate with high AI fluency over one with deep domain expertise.

That last figure represents a significant shift. And yet 59% of those same organisations report having made a bad AI hire – someone who sounded capable in the interview but couldn’t apply any AI skills once on the job.

One number explains most of that gap: 37% of organisations set their minimum bar for AI fluency at tool awareness. Meaning, knowing that a product like Copilot or ChatGPT exists. Describing it well enough in an interview. That is not AI competency. And the cost of that confusion will be inherited by this new apprenticeship once it meets the open market. Employers who struggle to articulate what they want from a candidate will face the same problem when they take on an apprentice. Providers who built tool-led digital syllabi will be asked to build something more behavioural. And an end-point assessment culture that already over-relies on surface signals will be asked to assess a skill that interviews are demonstrably bad at catching.

What this standard gets right

Skills England is trying to avoid what I’d call the confidence-versus-competence problem. The occupational profile describes entrants learning to identify where AI and automation can save time, reduce costs, and improve performance – and to integrate systems that do not currently talk to each other. That is workflow thinking, not just tool-naming.

Phil Smith, chair of Skills England, has framed the apprenticeship as delivering productivity gains in a responsible way, and the government’s wider positioning of AI as part of a fourth industrial revolution signals this standard is meant to sit within a broader economic strategy, not stand alone as a software certification.

That puts it squarely in the part of the economy where AI fluency is least understood and most consequential – the small-team marketer, the operational lead at a charity, the owner of an SME who has never had a data team. These are exactly the workplaces where the difference between AI-aware and AI-fluent shows up first, because there is no large IT function to absorb the cost of a poorly designed apprenticeship.

Where the standard could go further

The apprenticeship covers responsible AI use, but frames it primarily as compliance: protecting sensitive data, avoiding bias, meeting regulatory requirements. That’s necessary. But it’s not what employers tell me they’re struggling with.

When I asked hiring managers to rank dimensions of AI fluency by importance, responsible and ethical AI use came out at the top – ahead of applied AI use. The reason is operational rather than philosophical. If a team member uses AI in a way that mishandles sensitive data, or surfaces a result they cannot verify and don’t flag, that’s not just an embarrassment: it’s a liability. 

The standard’s current framing teaches an apprentice the rules. It should go further and teach them when to apply judgment in the gaps the rules don’t cover: when to disclose AI use, when to escalate a result, when to override a recommendation that looks plausible but feels wrong.

The same applies to what I’d describe as human-AI collaboration – the ability to use AI in a way that others on a team can audit, learn from, and build on. This is not a soft skill. It’s the dimension that determines whether an organisation gets the compounding benefit of AI-fluent staff, or simply accumulates individuals who use AI in isolation and create knowledge silos around themselves. When I asked employers what they found hardest to define, 54% pointed to making AI fluency work consistently across technical and non-technical roles. That gap is exactly what a well-designed apprenticeship could close.

The management question

The withdrawal of the management apprenticeship suite, set against the launch of a new AI practitioner standard, has prompted a fair question across the sector: is the Level 4 AI and Automation Practitioner essentially the new management pathway?

It is not a like-for-like replacement. Management involves more than AI fluency, and a Level 4 technical apprenticeship is not designed to develop the breadth of leadership and accountability that the Chartered Manager standard did at Level 6. But as the foundation of a redesigned management pathway, the logic is stronger than it first appears.

The job of a manager has changed faster than the qualifications used to describe it. A manager in 2026 is no longer just a manager of people – they’re managing hybrid human-AI workflows. Deciding where to automate and where to keep human judgment in the loop. Modelling responsible AI use for a team. Thinking through how one team’s use of AI ripples into another’s work. Those skills will need to be developed and accounted for in whatever comes next.

The risk in treating this AI standard as a management replacement is obvious – Level 4 is not Level 6. Forcing one to do the work of the other short-changes both the apprentice and the employer. But equally, if the AI standard is designed around behavioural fluency rather than tool training, it could produce exactly the foundation a future management pathway needs to build on. Get it wrong, and there will be two qualification gaps to fix instead of one.

Three things worth building in now

Based on the hiring data I’ve seen, here are three things I’d want in place before the first cohort starts.

First, design end-point assessment around evidence, not interviews. A significant share of the organisations that made bad AI hires left assessment entirely to the discretion of individual hiring managers, with no shared rubric. That same vulnerability exists in apprenticeship end-point assessment if professional discussion is the dominant mode. Structured, scenario-based tasks scored against a consistent rubric will produce a more defensible signal than an interview-style format.

Second, build the responsible AI module around judgment, not just compliance. Apprentices need to know the rules – but they also need to know what to do when no rule covers the situation. When to disclose AI use. When to escalate a result they can’t verify. When to override a recommendation that looks plausible but shouldn’t be trusted without scrutiny.

Third, design this apprenticeship with one eye on whatever replaces the management pathway. The Level 4 standard will produce people who manage AI-augmented workflows. A reshaped management apprenticeship, whenever it arrives, will need to build on exactly that foundation. If the two standards are designed in separate rooms by separate awarding organisations with separate assessment frameworks, the sector will end up with two qualifications that fail to talk to each other – in exactly the way the workflows they’re meant to support already do.

Looking ahead

The hiring market has spent two years discovering that AI fluency is hard to define and even harder to test for. The FE sector now has the chance to develop apprentices whose capability sits in their judgment, not in the toolkit they happen to know today.

Get it right, and the new apprenticeship will produce exactly the kind of practitioner that too many employers are currently struggling to hire.

By Wouter Durville, CEO and co-founder of TestGorilla