Two senior roundtables on AI adoption kept returning to the same quiet risk: the tasks that once taught young people the job are being automated without anyone deciding to. Jenny Garrett OBE argues this is a leadership choice, and sets out what employers, FE and policymakers could do.
In a room of senior People leaders earlier this year, someone described their organisation’s talent pipeline in a single word. It is no longer a pyramid, they said. It is a barrel. Broad in the middle and hollow at both ends. No juniors learning the craft. No senior leaders who came up through the new kind of work. Just the middle, holding everything together.
I have not stopped thinking about that barrel. The hollow at the bottom has a number. Between April and June this year, the ONS counted 981,000 young people not in education, employment or training.
Since March, we at Jenny Garrett Global has convened two invitation-only roundtables of senior leaders, mostly Chief People Officers and HR Directors, from financial services, energy, the NHS, local government, education, media and the cultural sector. We asked about AI adoption, first through the lens of a two-tier workforce and then through trust. We did not set out to talk about young people. They kept coming up anyway.
What I heard, set against the published evidence, suggests something FE leaders and policymakers may want to hear directly from employers. The first rung of the ladder is not being removed by AI. It is being removed by decisions leaders are making about AI, often without realising they are making them.
The rung is going, and employers know it
The evidence is now hard to ignore. A Survation survey of 1,001 UK businesses for Lancaster University’s Work Foundation, reported in August, found 36% had cut entry-level roles in the past year, rising to 46% of large firms. Around six in ten of the largest employers said AI and automation had reduced the jobs available. Nearly three quarters said it is harder for young people to get a foothold than five years ago. Adzuna data puts UK entry-level vacancies down 32% since ChatGPT launched. Alan Milburn’s interim review described the first rung as “simply out of reach” for many young people.
Yet noone in our rooms described a decision to stop hiring young people. What they described was quieter. Junior tasks absorbed by tools layered on top of the day job. Intakes paused while the organisation works out what AI changes. In law, AI contract review means some firms no longer need the volume of junior lawyers they once did. Each choice is defensible on its own. Together, they hollow out the opportunities.
Here is the arithmetic leaders are not yet doing. If an organisation stops hiring juniors now, by 2030 it will be short of the mid-level people those juniors would have become. The efficiency gained today may be borrowed from the capability needed tomorrow.
Who gets the tools is not neutral
One story from our July roundtable has stayed with me. A senior People leader described rolling out AI across several thousand employees. The protected, licensed version, the one safe to use with organisational information, went to whoever put themselves forward.
The leader named the problem before I could. Who puts themselves forward? People who already feel entitled to ask. People with a manager who will say yes, the time to make the case and the confidence to risk refusal. Opt-in looks scrupulously fair. In practice, it may select for confidence and proximity to power.
The national data points the same way. Cornerstone OnDemand’s survey of 2,000 UK employees found 16% of those earning £15,000 or less receive formal AI training often or always, against 47% of those earning over £55,000.
For FE, this matters twice. The learners you serve are often the least likely to raise their hand in an opt-in process. And once they are in work, AI fluency may compound the advantages others already hold. That is not a skills gap. It is an access design choice, and it can be designed differently.
There is a hopeful version of this. Dr Pauldy Otermans, a neuroscientist whose Otermans Institute has trained more than 35,000 learners across 13 countries, did not start with AI. She started with a problem: young people from less affluent backgrounds arriving at work without the invisible curriculum of cover letters, interview practice and unwritten rules. Her AI runs mock interviews in private. “You’re not feeling judged. No one is reporting this to anyone,” she told us. It runs on a phone, even without internet. Her principle could guide every college: AI should be a rehearsal layer, not another gatekeeping mechanism.
The young people refusing AI might be right
This is the part of the conversation I did not expect. Several leaders described early-career people choosing not to use AI. One young marketing professional put it plainly: if I use AI for everything at the start of my career, where is my learning?
The easy reading is resistance, or a training need. One participant read it differently. They train early-career people in a specialist conservation craft that has all but disappeared from university courses. Their trainees were actively steering clear of AI. Rather than ask why, the participant asked a better question: what do you find difficult?
The answers were cataloguing, tagging, and searching for something you know exists but cannot find. None of that is the craft. All of it is why the craft runs out of time. The participant’s advice was to use the tool for the wireframe, the first structure and the general research, and then go deep yourself, the way you always would. That is where the learning stays.
That is not a technology policy. It is a pedagogy. And it is exactly what good FE already does: teach the craft, and teach which parts of the work are scaffolding.
There is the challenge worth sitting with The young people most sceptical of AI may be protecting the very judgement employers will need from them in ten years. An executive search participant offered the hopeful version. Organisations hold experienced people with deep knowledge and junior people fluent with the tools. AI could be the medium for transfer between them, rather than a replacement for one of them.
This is a leadership decision
I have spent twenty years in leadership development, and this looks to me less like a technology problem than a leadership one. Three decisions sit with the executive layer, and they cannot be delegated down.
- Decide which work stays human, and say so. Name the tasks where people learn their craft, and protect them deliberately. Ambiguity here is what makes early-career people anxious and what hollows the bench.
- Audit how access is actually allocated. Not what the policy says, but who the mechanism has selected. Then decide whether that map is a strategy or an accident.
- Answer the time question. When AI frees up time, where does it go? One answer that pays back is development: back into the apprentice, the graduate and the new starter.
The middle of the barrel cannot do this alone. Writing in Harvard Business Review in June, Julia Shin and Sandra Sucher found middle managers absorbing the validation, quality control and coaching that AI creates, with no extra support. If juniors are to learn alongside AI, someone has to design that learning on purpose.
What FE and policymakers could do
Colleges and training providers may be the best-placed brokers in this conversation, because they sit between employers and the young people those employers will need. Three questions might be worth putting to every employer partner this year:
- Which tasks have you automated that used to teach new starters the job?
- What has replaced them as a way of learning?
- Who in your organisation decided that, and do they know they did?
For policy, one gap may be worth naming as Milburn’s final recommendations take shape. We measure NEETs carefully. As far as I can find, nobody measures what happens to entry-level roles when an employer adopts AI: redesigned, retained or removed. When something is not measured, it tends to be decided by default. Perhaps large employers could be asked to report what happened to their entry-level roles alongside their AI adoption.
Back to the barrel
A barrel holds things. It does not grow anything. An organisation shaped like one can run efficiently for a while, but it is living on capability it is no longer building.
The first rung is not being taken by AI. It is being taken by leaders who have not yet noticed they are holding the saw. The good news is that a decision made by leaders can be remade by leaders.
So here is the question I would leave with every employer, principal and policymaker reading this: in your organisation, which tasks used to teach young people the job, and who is learning them now?
By Jenny Garrett OBE, Founder and CEO of Jenny Garrett Global