Teachers at Euan Blair’s £1.6bn tech training company, Multiverse, have blown the whistle about “horrendous stress” and feeling constantly watched after bosses started using AI models to surveil and rate their teaching.

Staff delivering apprenticeship training in computer and AI skills to thousands of UK public and private sector workers now have transcripts of online classroom sessions analysed and scored by an AI programmed to alert human managers when they are suspected of doing things wrong.

The AI is trained to flag problems in their performance ranging from not fixing a disruption to the online connection quickly enough – within about a minute – to using too many “filler” phrases, such as “sort of” and “kind of”.

Teachers called the newly introduced system “remorseless” and “unnerving”, and said some staff were on edge, losing sleep, and seeking therapy.

Multiverse was co-founded by the eldest son of the former prime minister Tony Blair and employs more than 800 people. It supplies AI and digital skills training to workers in the NHS, local councils, universities and the private sector, mostly funded through the government’s apprenticeship levy.

Alongside his father, Blair is one of the biggest cheerleaders for wider AI adoption in the UK economy, claiming last week that “every single worker is going to need to be retrained” and that employers want their workforces “as AI-pilled as possible”. Multiverse uses a mix of AI models including Anthropic and OpenAI products.

Its new AI teacher monitoring system tags each instructor by their “risk status”, calculates a “percentage confidence score” and produces written comments about teachers’ work for managers to see at a glance.

It is only the latest example of employers using AI to track worker behaviour. Earlier this year Burger King started using AI to track restaurant workers’ customer interactions and this summer Meta paused tracking of employees’ computer keystrokes amid data privacy concerns and a staff backlash.

With critics warning AI monitoring risks eroding human agency, one study has suggested it can backfire by turning workers rebellious and new research to be published this week reports that workers in the wider UK tech industry are finding AI degrades work satisfaction. The United Tech and Allied Workers branch of the Communication Workers Union concluded AI causes “cognitive surrender” as workers lose skills and confidence and often intensifies rather than eases work.

“Since all of this has come in, the stress level has been absolutely horrendous,” said one of several Multiverse teachers who spoke on condition of anonymity. “It has affected sleep, and not just for me. I’ve gone through a lot of things in my life, and I haven’t ever been in a position where I’m waking up in the middle of the night.”

Previously instructors were observed monthly by a human manager but are now watched for several hours per day by an AI. It is trained to consider marking instructors down if they give vague or uncertain answers to questions, if learners go quiet when asked to contribute and if the teacher gets pulled into a prolonged one-to-one exchange that sidelines the group.

“This is inducing stress because you consistently have to be cautious,” one teacher said. “Your attention is no longer delivering [the lesson] and listening to the learner [but] am I saying the right thing so the AI doesn’t mark me down for something stupid that I’ve mistakenly said.”

Another teacher called it “very uncomfortable”.

Multiverse says the AI does not replace human judgment but “directs human time to where it’s needed most” by highlighting problem. A spokesperson said it was “using AI to improve the experience of learners and customers”, and that only human managers wrote teachers’ performance reviews.

They did not directly address complaints the system caused stress and anxiety and felt “remorseless”, but said its introduction was communicated in advance and teachers will have access to the data the AI collects.

“This agent provides instructors with improved advice on development and feedback – exactly how any learning organisation should be deploying AI,” the spokesperson said. “The most consequential performance management action it can take is to recommend that a human personally reviews a session.”

Blair owns a large minority stake in Multiverse, putting his personal net worth at an estimated £350m. The company was recently valued at £1.6bn albeit on a turnover of £118m and losses of £28m last year. In February, Ofsted, the UK education inspectorate, found Multiverse failed to meet expected standards in several areas, but praised the instructors as “skilled, effective teachers”.

Its new AI monitoring agent, introduced in recent weeks, is trained to spot signs an instructor has not fully internalised the teaching material. The AI is instructed to count verbal tics that point to that conclusion and to “judge the PATTERN across the whole transcript, not isolated instances”.

It says: “1-2 hedges or filler words in an otherwise direct, well-internalised explanation do NOT fail this standard … What fails the standard is a PERVASIVE REPEATED pattern of filler, hedging or re-explanation.”

When a trainer is setting a task “if the instructor is visibly confused themselves or had to self-correct while giving the instruction, that confusion itself counts as evidence the instruction was not clear the first time”. The AI is told to apply a negative mark “even if it was eventually straightened out”.

The workers said the AI monitoring has led to managers pulling them up for suspected “digressions” when in fact they were answering a student’s question.

“It’s impossible for me not to be affected by this mentally,” one teacher said.

“I feel like I’m under continuous scrutiny,” another said. “It’s all about what I didn’t do in this session rather than what I did do. Did I do something incorrectly yesterday? What will I mess up on today? It’s remorseless. I don’t know why this is happening.”

John Chadfield, national tech officer at the CWU, said the public would be horrified to hear of bosses “deploying deeply intrusive surveillance on workers” but said “this news won’t surprise tech workers who feel the reality of the tech sector – an industry in serious need of further regulation.”

He called for new laws to ensure AI is deployed in a socially responsible way and not to “shut off job opportunities for young people, degrade the craft of skilled workers and make astronomical profit for a few”.

Case study: ‘An AI judges my lessons. Multiverse calls it quality assurance – it feels like surveillance’

Being constantly monitored by an AI is deeply uncomfortable. Imagine every word you say on the job being checked against a rubric that dictates what you’re expected to say and how. That’s the reality for me and my colleagues: every session we deliver to apprentices is now recorded and transcribed. It feels intrusive. Multiverse calls it quality assurance, but it feels like surveillance.

Managers can now look at a dashboard showing how the AI has judged each of our sessions. It assigns a risk level to each of us: the risk of not having said the right things. That’s stressful, for several reasons.

First, the AI isn’t always accurate. Managers are now tasked to investigate trainers for not following the script – at least according to the AI – when actually they were adapting the lesson to a learner’s needs. The threat of being questioned in such detail, at any time, is a constant source of anxiety.

Second, there is no transparency. During the rollout, all delivery staff were kept in the dark while the system was developed. We have no insight into the AI’s analysis. The company says staff will get dashboard access in time, but for now only managers can see it.

The need to always be cautious warps your priorities. Instead of focusing attention on the learner, you become obsessed with whether you’re saying the right thing. Being on camera for most of the day is stressful in itself; being haunted by an AI makes it worse.

A more productive way to use AI would be to improve the teaching materials and raise the standard of what we teach. Instead, the AI is checking whether the teacher managed to fill the gaps? As it stands, this is surveillance, not empowerment.

And I’m worried about what this data is for. In a climate of AI replacing people, I can’t help but worry it will be used to train AI to replace coaches and instructors altogether. Apprentices don’t want that and neither do I, and I would never consent to my sessions being used this way.