Instructors delivering digital and artificial intelligence apprenticeships at Multiverse, the £1.6 billion training firm founded by Euan Blair, have spoken out about extreme stress and a pervasive sense of being watched after the company introduced an artificial intelligence system to monitor and score their classroom performance.
The AI‑Powered Monitoring Regime
The new system tags each trainer with a “risk status” and calculates a percentage confidence score based on analysis of transcribed session recordings. It flags behaviours such as prolonged pauses, excessive use of filler phrases like “sort of” or “kind of”, and any deviation from a preset script. Human managers then receive written comments that summarise the AI’s findings, which can influence performance reviews. Multiverse says the tool is intended to direct human attention to areas that need improvement, but the reality for staff is a constant audit of every word spoken.
According to the developers, the algorithm is trained on a mixture of models, including those from Anthropic and OpenAI, and is instructed to consider the overall pattern of speech rather than isolated incidents. It counts verbal tics that suggest a lack of mastery and will assign a negative mark even if the teacher later corrects a confusion during the lesson. The system can also flag a “digression” when a trainer answers a learner’s question, treating the extended exchange as evidence of poor lesson focus.
Employee Reactions and Psychological Toll
One anonymous teacher described the atmosphere as “horrendous” and said the pressure had robbed him of sleep, a condition he had never faced before. “Since the system arrived, the stress level has been absolutely horrendous,” he said, adding that the anxiety had spread to his colleagues, some of whom are now seeking professional help.
Another instructor explained that the constant scrutiny forces staff to second‑guess every sentence, shifting focus from the learner to the risk of being penalised for a stray phrase. “You’re no longer teaching; you’re worrying whether the AI will mark you down for something stupid you said,” she observed, noting that the feeling was “very uncomfortable”.
A single‑sentence emphasis: the monitoring feels like surveillance, not quality assurance.
Multiverse maintains that the AI does not replace human judgment; it merely highlights issues so that managers can intervene where needed. A spokesperson said the system improves the learner experience and that only human managers write performance reviews, though the company has not directly addressed the reported stress and anxiety.
Regulatory Oversight and Union Critique
Ofsted, the UK’s education inspectorate, recently judged Multiverse as failing to meet several standards, though it praised the instructors as “skilled, effective teachers”. The agency has not commented on the AI monitoring tool, but its findings add pressure on the company to demonstrate that its practices do not compromise educational quality.
John Chadfield, national tech officer for the Communication Workers Union, warned that the public would be shocked to learn that bosses are deploying such intrusive surveillance. He called for new legislation to ensure AI is used responsibly, arguing that the technology should not undermine skilled work or concentrate profit in the hands of a few.
Earlier this year, Burger King rolled out AI to track restaurant staff interactions with customers, and Meta paused monitoring of employee keystrokes amid privacy concerns and staff backlash. These precedents illustrate a growing trend of employers using data‑driven tools to police worker behaviour, raising broader questions about labour rights in the digital age.
Implications for the Broader Use of AI in Workplace Training
The Multiverse case highlights how AI can be repurposed from a tool of empowerment to a mechanism of control. While the company argues that the system points to genuine developmental needs, critics contend that the opaque algorithm erodes trust and may ultimately lead to the replacement of human trainers by automated feedback systems.
Experts suggest that transparent, participatory design processes and clear limits on data use are essential to prevent the kind of resentment that is now simmering among Multiverse staff. Without such safeguards, the technology risks turning learning environments into high‑pressure arenas where educators are judged more on compliance than competence.
Why it Matters: The way Multiverse handles AI surveillance will set a precedent for how UK organisations balance innovation with employee wellbeing, influencing future policy, public trust, and the long‑term viability of AI‑enhanced training programmes across the economy.
