TL;DR
Measuring staff on how well they use AI has moved from novelty to policy at a growing number of large employers, with bonuses, promotions and in some cases continued employment now attached to it. UK employment lawyers say the practice is lawful. What it is not is settled: the productivity a worker unlocks becomes the new baseline, without a corresponding change to their pay, and some of the earliest adopters have already retreated.
The bargain workers are being offered
The BBC’s account opens with a marketing employee whose bonus depends on getting more done through AI, and who has worked out the arithmetic. Clearing an extra two days of work in the time saved buys him neither a lighter week nor a proportionate rise; it simply resets what is expected, while demonstrating how much of the role no longer needs a person in it.
That tension is spreading. Accenture’s chief executive has said plainly that AI is now how work gets done there, and advancement follows. Internal leaderboards ranking employee tool use have appeared at KPMG, JP Morgan, Meta and Disney, according to reports cited by the broadcaster, while Coinbase dismissed engineers who did not complete mandated AI training.
A senior executive at a large consultancy, speaking anonymously, described a two-tier workforce forming — identical titles and tenure, different standing — and put it starkly: fluency now outranks credentials, so someone with three years and quick hands gets past someone with twenty and none.
Where UK law sits
Tina Chander, an employment partner at Weightmans, confirmed employers are within their rights, then posed the awkward follow-on: when AI makes someone more productive, does their reward change to match — or have they quietly engineered their own redundancy? Her advice is explicit policies and boundaries, because the alternative is disputes over fairness and job security.
Timing matters here. Unfair dismissal claims in the UK become available after six months of service rather than two years from January 2027, and the window for lodging one doubles from three months to six. Vague AI expectations become considerably more expensive against that backdrop.
Looking forward
Two large employers have already reversed. Duolingo dropped AI use from performance reviews after staff asked whether the point was AI for its own sake, and Amazon shut an internal leaderboard when people began inventing tasks to climb it. Henley Business School’s Kamila Miller put the failure mode neatly: make usage a KPI and you measure obedience, not judgement. Set against research that half of workers think AI makes them look better than they are, UK firms writing this into review criteria should be clear which of the two they are actually rewarding.