What the Co-op AI backlash teaches SMEs about staff trust

The technology worked. That is what makes the Co-op Legal Services story so instructive.

In October 2026, the Guardian reported that Co-op Legal Services had deployed an AI system to record, analyse, and automatically grade staff phone calls with probate clients. A whistleblower described the system as oppressive. Co-op Legal Services rejected that description, saying the system helps it maintain standards and coach staff, and that it had received no formal complaints. But the story shows how quickly trust becomes the issue. None of it was about the AI malfunctioning. It was about whether staff felt like participants in the deployment or subjects of it.

For UK HR Managers, Operations Managers, and Managing Directors weighing up AI quality-assurance tools, this case is not a reason to avoid the technology. It is a precise map of where adoption goes wrong.

What Co-op Legal Services actually deployed

The system automatically monitored phone conversations between legal advisers and bereaved clients seeking probate guidance. It scored each interaction against predefined performance criteria. Scores were used to assess individual staff members.

The capability itself is not unusual. AI speech-analytics tools have been in commercial use across financial services and contact centres for several years. They can flag compliance risks, identify coaching opportunities, and surface quality trends across large call volumes far faster than manual review.

What was unusual was the manner of deployment. According to the whistleblower account reported by the Guardian, staff were not meaningfully consulted before the system went live. The scoring methodology was not transparently explained. And the outputs fed directly into performance management, without a human-led review stage acting as a check on the automated judgement.

Why this matters more than a PR problem

It would be easy to read this as a communications failure. Brief the team better, update the intranet, hold a town hall. Problem solved.

That reading misses the point. If the whistleblower account is accurate, staff were not involved in designing the system that would judge them. That is not a communications gap. It is a governance gap.

When automated monitoring arrives without consultation, several things happen simultaneously. Trust in management falls. Union or employee representative bodies mobilise. Staff begin to modify their behaviour in response to the scoring criteria rather than client need. And the quality data the system was designed to generate becomes less reliable, because the behaviour it is measuring has already been shaped by awareness of being measured.

For operations managers, that last point is the practical one. A monitoring system that changes the behaviour it is trying to measure does not give you an accurate picture. It gives you a performance theatre picture.

The three risks UK SMEs underestimate

Most UK SMEs adopting AI monitoring tools focus their risk assessment on the technology itself. Does it integrate with the phone system? Is the vendor reputable? What does the contract say about data ownership?

These are reasonable questions. They are also the wrong starting point.

The risks that actually derail AI adoption in this space sit elsewhere.

The first is cultural rejection. Staff who feel surveilled without consent withdraw. They comply with the letter of the scoring criteria and disengage from the spirit of the work. For a firm like Co-op Legal Services, where advisers are speaking with recently bereaved clients, that disengagement has a direct client-experience consequence. For any SME where staff judgement matters, the same dynamic applies.

The second is legal exposure. Automated performance monitoring touches employment law, data protection obligations, and in unionised environments, collective bargaining agreements. A Data Protection Impact Assessment is not optional for systems of this type. It is a legal requirement under UK GDPR where processing is likely to result in high risk to individuals. Deploying without one is not a grey area.

The third is reputational damage. The Co-op story reached a national newspaper. For a smaller organisation, a similar story reaching local trade press or Glassdoor can have outsized recruitment and retention consequences at a moment when hiring is already difficult.

HR Managers need to be in the room when these tools are being scoped, not brought in after the vendor contract is signed.

What responsible deployment actually looks like

The answer is not to avoid AI quality-assurance tools. Used well, they genuinely help. They can identify coaching opportunities that manual spot-checks miss. They can surface compliance risks before they become regulatory issues. They can reduce the burden on supervisors who currently spend hours listening back to recorded calls.

The answer is to change the sequence.

1. Conduct a Data Protection Impact Assessment before any procurement decision. This means mapping what data is processed, who has access to scores, how long data is retained, and what decisions the scores inform. If a DPIA is new territory for your organisation, that is a signal to pause and get structured advice before proceeding.

2. Involve staff representatives in designing the feedback loop. This does not mean giving staff a veto over every design choice. It means explaining what the system measures, why those criteria were chosen, how scores will be used, and what recourse exists if a staff member believes a score is inaccurate. Transparency reduces fear. Fear is what drives resistance.

3. Build in a human review stage. Automated scores should inform a conversation, not conclude one. A managing director or operations manager who treats an AI grade as a final performance verdict has outsourced a judgement call that should stay with a human. The technology flags. A person decides.

4. Pilot before you scale. Run the system in a single team or function for a defined period. Measure not just the quality data it generates, but the effect on staff sentiment. Then adjust before rolling out further.

None of these steps are technically complex. All of them require time and organisational honesty that is easy to skip when a vendor is promising faster ROI.

The limitation nobody wants to name

AI monitoring tools are better at measuring what is easy to measure than what actually matters.

A call-scoring system can assess whether a staff member used approved language, covered mandatory compliance points, and kept the call within a target duration. It cannot reliably assess whether a bereaved client felt heard. It cannot score the moment a skilled adviser decided to go off-script because they recognised the client needed something different.

For customer-facing roles in professional services, that gap between measurable and meaningful is significant. Operations managers should be explicit with their teams about what the system does and does not capture. Otherwise staff will optimise for the metric, and the human quality the organisation actually cares about will quietly erode.

This is not an argument against the technology. It is an argument for using it as one input among several, rather than as a single source of performance truth.

Where to take this next

AI adoption is 80% people and culture, 20% technology. The Co-op story, as reported, shows what happens when an organisation appears to invert that ratio.

If your organisation is considering any AI tool that touches employee activity, whether call monitoring, productivity tracking, or workflow analysis, the first question is not which platform to choose. It is whether your team understands what is being deployed and why.

gecco's Training and consultancy programme helps SME teams understand AI tools before deployment, reducing the risk of staff distrust and regulatory exposure. By grounding adoption in the SCALE model, gecco helps organisations scope AI use responsibly, build internal competence, and apply tools in ways staff understand and accept.

If you are an HR Manager, Operations Manager, or Managing Director thinking through where AI quality assurance fits in your organisation, and what responsible scoping actually involves, the AI Readiness survey is built to surface exactly those questions. Taking the survey gives you access to 65+ free resources and a custom AI Readiness report, followed by a free 30-minute showcase to walk through the results with a member of the gecco team.

Website · LinkedIn · Case Studies · Newsletter