TL;DR

Co-op Legal Services has started using AI to record and grade the phone calls its advisers handle on probate, wills and estates, the Guardian reported on Thursday 8 October. An OpenAI model checks over 50 separate elements of each call and gives managers pass and fail marks. A worker described the system as “oppressive and dystopian”; the Co-op says it is a support tool, not a decision-maker.

What the system does

Each conversation between an adviser and a customer gets a percentage score, with the model’s individual pass or fail judgements fed to managers for performance analysis. Many of those callers have recently been bereaved. The monitoring reportedly arrived in the past couple of months, covers dozens of staff, and is partly used to find ways to lift their sales.

The anonymous worker told the Guardian that being reviewed now and then for compliance felt very different from having every word assessed, sometimes for several hours a day. “You are being watched,” they said, adding that it bred distrust.

Caoilionn Hurley, who runs Co-op Life Services, said AI lets the business review all customer conversations and supports coaching for colleagues, while “human judgment, accountability, care and kindness” stay central. The Co-op rejected the criticism, saying its probate advisers were highly engaged.

A wider pattern

The Co-op is not alone. The Guardian reported a week earlier that Multiverse, the training company founded by Euan Blair, was using AI to monitor its online classes. Burger King announced similar tracking of US restaurant staff in February, and Meta paused logging of employee keystrokes this summer after objections.

John Chadfield, national officer at the Communication Workers Union, argued that “unaccountable computer systems should not be people’s managers”. The International Labour Organization has also warned this year about the stress risks of AI monitoring.

Looking forward

For UK employers, this is a test case for how far automated quality assurance can go before staff treat it as surveillance. In our view, the claim that AI is “not a decision-maker” is hard to square with scores going straight to managers for performance data, and with its partial use to improve sales performance. Firms weighing similar tools would be wise to tell staff exactly which decisions the scores can and cannot inform, before a whistleblower does it for them.