UK's Co-op Becomes the Latest Example of AI Surveillance of Employees at Work

Co-op, the UK consumer cooperative that runs thousands of businesses, is the latest example of AI systems watching over workers. The Guardian reports this in its TechScape newsletter, where author Blake Montgomery says algorithms are tracking employees around the world more and more often. The available excerpt of the piece contains no technical details or figures about Co-op itself, so below we look at the context rather than retell the inner workings of a specific system.

What we know about the Co-op case

The published announcement supports three facts. Co-op is a large British cooperative with thousands of outlets and business lines. It is described as the latest in a series of examples of employers using AI to monitor people. The topic is framed as a global trend, not as the local story of a single company.

Scale matters here. When an organization with thousands of units launches monitoring, the system immediately runs across a huge number of shifts, checkouts and warehouses. A classifier error or a questionable performance metric turns into thousands of decisions about bonuses, schedules and disciplinary action.

Where the surveillance market comes from

Monitoring tools are not new. Warehouses and call centers have lived with timers and task counters for decades. The shift to remote work added so-called bossware: software that logs computer activity, keystrokes and time spent in applications, and sometimes takes screenshots. Generative AI has added a new layer. Models can analyze not only numbers but also the text of correspondence, calls and video, and then assemble an assessment of an employee from them.

A typical set of signals these systems collect:

• work pace: operations per hour, idle time, break length;
• device activity: applications, websites, keyboard input;
• audio and video: conversations with customers, camera footage from sales floors;
• text channels: work chats and email;
• derived scores: ratings, "attrition risk," "engagement."

The further these systems move from direct counters and the closer they get to "mood" and "loyalty," the weaker the scientific basis for their conclusions and the higher the risk of error.

The legal framework

In the UK, employee monitoring is governed by UK GDPR. The regulator, the ICO, has published guidance: monitoring must be proportionate, and employees must be told in advance, and in clear terms, what is collected and why. In the European Union, the AI Act is stricter: since February 2025, emotion recognition in the workplace has been banned, with narrow exceptions. For companies operating in several jurisdictions, this means the same system may be acceptable in one country and problematic in another.

Automated decision-making is a separate risk area. If an algorithmic rating affects dismissal or pay, the person must be able to learn the logic behind the assessment and challenge it. In practice, such channels are often either undocumented or work only on paper.

What employees and employers should do

For employees, the basic step is simple: request a written description of what data is collected, where it is stored and how it is used in evaluations. In the UK and the EU, people have a right of access to their personal data for this purpose. Work devices and corporate accounts are best treated as an observed environment and kept separate from personal ones.

For employers, a practical minimum looks like this: a data protection impact assessment before launch, a clear purpose for collection (security, service quality, not "everything at once"), a named person responsible for disputed decisions, and regular checks of the metrics for bias. Unions and worker representatives are better brought in when the system is being chosen, not after the first complaints.

The Co-op case, judging by the Guardian's announcement, will prompt new questions for retail. Full details of the system will appear in the publication's main article, and they will show where the line falls between quality control and total surveillance.