Co-op Legal Services puts every probate adviser's call through AI grading

For the advisers involved, what has changed is how much of their work is listened to. "It's different from occasionally being listened to for compliance purposes," a Co-op worker told the Guardian on condition of anonymity ^[9]. Caoilionn Hurley, managing director of Co-op Life Services, said full coverage was the point. "The efficiency of AI over traditional methods allows us to review all customer conversations to ensure we consistently deliver high standards of service, while enabling more informed support, coaching and development for colleagues," Hurley said ^[8]. The worker put it more bluntly: "You are being watched" ^[9].

The trade-off sits inside a single score. Co-op said the AI helps ensure "our clients consistently receive empathetic expert guidance" ^[17]. The same monitoring is partly there to find sales ^[5], on calls that one worker said often come from the recently bereaved ^[4]. On a probate call, empathy and selling can pull against each other. The report does not say who set the 50-plus criteria or how they were tested against calls of that kind ^[3].

Co-op's reply is the case a sceptic of the coverage would make, and it deserves some weight ^[6]. The staff criticism in the report comes from anonymous sources, and the system covers dozens of legal services staff ^[10][7]. The weaker part of the reply is Co-op's statement that the tool is not a decision-maker. Managers use its pass and fail scores when they analyse employee performance ^[2], so the tool shapes the judgement even when a person makes the final call.

The wider claim, that blanket scoring is becoming routine, has limited support in the reporting so far. Last week the Guardian reported that Multiverse, an AI training company, had started AI monitoring of online classroom sessions. Teachers called it "remorseless" and "unnerving" ^[11]. Burger King announced in February that it would track US restaurant workers' customer interactions ^[12]. This summer Meta paused tracking of employees' keystrokes amid data privacy concerns and a staff backlash ^[13]. The coverage names four employers, and one of them has stepped back ^[20].

Staff wellbeing takes the cost first. This year the International Labour Organization warned of the "psychosocial" risks of AI monitoring, including work-related stress ^[15]. "AI systems are being deployed in workplaces in the absence of a comprehensive regulatory framework," it said ^[16]. John Chadfield, national officer of the Communication Workers Union, said: "Unaccountable computer systems should not be people's managers, and these hard-working employees demand far more respect" ^[14]. Hayfa Mohdzaini, a senior policy and practice adviser at the Chartered Institute of Professional Development, was more neutral. "It will be interesting to see how it develops and how much managers need to adjust their approach," Mohdzaini said ^[18].

For any employer running a system like this, the decision this quarter is where the pass and fail marks are allowed to go. If they stay inside coaching, they fit Co-op's description of the AI as a support tool ^[6]. If they go into performance ratings or sales targets, the CWU's objection to computers acting as managers becomes harder to answer next quarter ^[14]. I think Co-op's current design is closer to the second path. The marks already feed managers' performance analysis, and part of the system's purpose is sales ^[2][5].