More than a million young people aged 16 to 24 in the UK are not in employment, education or training, and projections suggest 1.25 million within five years. The government is trialling the use of AI as a tool to support people into work. It is one of the first public deployments of this scale targeting a vulnerable population.
We have covered extensively what AI eliminates in terms of jobs. Here is an attempt to use it in the other direction, and it deserves to be looked at without cynicism or enthusiasm.
What AI can actually bring here
The problem these young people face is rarely a lack of opportunities. It is a set of accumulated obstacles: not knowing where to start, not having the codes of a sector, not daring to apply, not knowing how to articulate an experience that seems insignificant.
On these specific points, the tool has demonstrable utility, and we described it in our article on recruitment: reformulating experience, tailoring an application, understanding what a job posting expects. These are formal barriers that have nothing to do with competence, and they penalise first those who did not get the right codes at home.
Add to that availability. An adviser sees someone once a month. A tool available at any hour, without an appointment or judgement, suits a population that has often given up on asking for help.
The three serious risks
The replacement of human support. This is the main risk, and it is budgetary before it is technological. An AI system costs significantly less than an adviser. The temptation to substitute rather than complement will exist, and it will hit precisely those who most need a relationship. A young person in crisis does not just need information: they need someone who notices they did not turn up.
The invisible error. Incorrect information about a right, a benefit or a procedure can be costly, and this population often has no one to cross-check it. This is the paradox we noted regarding access to justice: those who benefit most from the tool are those least able to verify its outputs.
The arms race. If all candidates produce assisted applications, the advantage disappears and the volume increases. We have described this mechanism: everyone runs faster just to stay in the same place.
A simple criterion allows judgement: does the system train or do it for you? A tool that writes the letter produces a document. A tool that helps prepare for an interview, that makes you practise, that asks the difficult questions, builds a capability that will remain. This is exactly the distinction we made in our article on schools, and it applies here with a more immediate stake.
What to watch
Three indicators will tell whether the trial has succeeded, and none of them is the number of applications sent.
The sustained exit rate into employment or training, measured at six months rather than at the end of the programme. The maintenance of human contact, that is, whether advisers have been strengthened or replaced. And the effect on those who drop out, the most disengaged population being the one that digital systems reach least well.
What to take away
It would be easy to mock the idea that a chatbot would solve a deep social problem. That would be unfair: no one claims that, and the tool can genuinely remove concrete obstacles.
It would be just as easy to welcome it as modernisation. That would be naive: youth unemployment does not come from a lack of tools, it comes from a set of economic, educational and social factors that no technology corrects.
The right position is probably this: a good tool serving strengthened human support can genuinely help. The same tool presented as an alternative to that support would worsen the situation by creating the illusion of action. The difference between the two is not visible in the technology; it is visible in the budget.