If AI does the junior work, who learns to do the senior work?

The UK has just passed the most significant overhaul of employment law in a generation. The Employment Rights Act 2025 will, among many changes, bring in a shorter qualifying period for unfair dismissal, remove the unfair dismissal compensation cap, and force employers to take ‘all reasonable steps’ to prevent sexual harassment.

Despite this wide-ranging scope, it fails to tackle one of the central issues in employment at present – automation.

Last week, Rishi Sunak, who is now an adviser to AI firm Anthropic and Microsoft, said AI is flattening the jobs market for young people, especially in service sectors such as law, accountancy and the creative industries.

AI is evidently displacing more of the work which had previously allowed young staff to learn their craft – for prospective lawyers that work would have included producing first drafts of documents, carrying out legal research, and drafting standard correspondence. When that junior work disappears and is swallowed up by the use of LLMs or AI agents, training grounds also disappear.

Employers, of course, are not blind to the economics. An AI licence costs a fraction of a graduate salary, requires no annual leave and doesn’t carry the risk of an individual bringing an employment tribunal claim. This has meant that junior intakes are being reduced, trainee programmes streamlined, and apprentice headcounts trimmed. Whilst some popular commentary has focused on jobs being replaced by someone using AI rather than being replaced by AI itself, junior employees are seeing the very work they need to do to enable them to learn their trade simply disappearing.

There is a material difference between learning to review work created by AI and learning to do it yourself. A generation trained only to supervise AI may never acquire the underlying knowledge needed to robustly challenge outputs.

The use of AI is a commercial decision for an employer, but the consequences of blindly doing so are already starting to show. The senior-level judgment needed to supervise and correct AI has to come from somewhere – from employees who gain the seniority to be able to do so by doing the very work which has now been automated away.

The Government, UK Boards and HR directors need to consider this risk hidden below the AI iceberg as a priority. It is a serious long-term issue, yet the cost of getting it wrong will not appear in this year’s P&L. That cost will appear in five or ten years’ time,  by which time the cupboard of next-generation senior decision-makers will be bare.

In the meantime younger graduates and applicants are competing for fewer and fewer entry-level roles, which involve less and less of the formative junior work they need to do to learn and progress in their careers and are faced with employers which can increasingly argue that the same output can readily be produced more cheaply with the use of AI.

None of this is an argument against AI. It is an argument against pretending that the displacement of junior work is somebody else’s problem, or that it will sort itself out in time. Employers should think very carefully now before treating their junior pipeline as a dispensable nicety.  The Government needs to consider now what steps it can take to ensure that all sectors are providing junior entry level career and learning opportunities so that there will also be a pipeline of successors in the future.

This is not about succession planning – it is about whether there will be any human successors.