Bossware Is Already Scoring Workers—41% Say Nobody Told Them

An October 8, 2026 report from the UC Berkeley Labor Center and partner groups describes digital systems that measure performance, assign work and schedules, and inform decisions about pay, discipline and termination; it also cites research finding that 41% of workers received no communication from their employer about workplace data collection or AI use. The crucial step is what happens after information is collected: an activity record can become an assessment that affects a person’s job.

An October 13 Guardian account describes Co-op Legal Services in the UK recording and analyzing customer calls and giving advisers percentage scores; Co-op says the technology supports staff and service quality. The case makes the process tangible. A customer conversation becomes structured performance information that a manager can use, while the adviser may have little insight into how the assessment was produced.

At Co-op, a customer call becomes a performance score

In the Guardian’s investigation of the call system, an anonymous worker said, “You are being watched”; the account describes a model assessing more than 50 aspects of each call and sending pass-or-fail results and percentage scores to managers. The calls concern probate, wills and estates, so an adviser’s performance is being assessed during conversations that can be sensitive for customers.

The worker described continuous analysis of conversations as a marked change from occasional compliance reviews. Co-op described the technology as a support tool that helps managers coach colleagues and maintain service quality, with human judgment remaining part of its process. Those accounts establish that calls are analyzed and scored, and that managers receive the results. They do not establish that a particular score caused a pay cut or dismissal.

The assessment changes what a manager can readily compare. Listening to an entire conversation takes time; a compact score can be placed alongside other advisers’ results. That makes the scoring criteria consequential: a difficult customer interaction, an interruption or a misunderstood phrase could affect the output unless the review process captures the surrounding circumstances. The open question for an affected adviser is which parts of the call counted and how a disputed assessment is corrected.

From activity records to shifts, pay and discipline

Bossware encompasses several different paths from observation to a job decision. Location records and task-completion times can inform assignments, routes and expected pace. Digital scheduling tools can allocate shifts and hours; timekeeping systems affect the hours recorded for pay. Productivity measures, alerts and rankings can also enter performance reviews or disciplinary decisions. These are documented categories of use, not a claim that every employer connects every measure to every outcome.

The weakness of a narrow measure becomes visible when it stands in for the whole job. A delivery stop can take longer because of traffic; a package scan can fail even when a parcel was handled correctly. Workers in the California research described disputed driving alerts, scanning errors and delays outside their control. In Los Angeles, freight drivers received automated termination notices after an algorithmic ranking evaluated them. That example shows the employment consequence clearly: a ranking did more than describe activity.

There is a difference between data a tool collects, an alert or score it produces, and the decision made from that output. A location record might help dispatch work; the same record could be used to judge whether a driver kept pace. A manager may weigh an alert against a worker’s explanation, or a workplace process may give the alert greater weight. Knowing that tracking exists therefore does not, by itself, reveal how much influence its output has over hours, pay or discipline.

What workers are told about monitoring

The disclosure finding concerns communication about data collection or AI use; it does not establish that all workers who heard nothing were secretly given performance scores. Parriva’s account of the California research explains that its 25 worker interviews document experiences rather than measuring how common monitoring or automated discipline is across the state. The broader disclosure figure comes from earlier research cited in the report, not from those interviews.

The interviewed workers’ experiences show why a notice about a device may be too vague to explain its consequences. Some learned of monitoring through colleagues, an accidental app notification or a feature encountered by chance. A driver may know a vehicle records its location without knowing whether stops affect a ranking. An adviser may know calls are recorded without knowing which features of a conversation are assessed or who sees the resulting score.

The useful questions follow the path from record to decision: What information is collected during a shift? Which alerts, rankings or scores are produced from it? Can those outputs affect hours, pay, evaluation or discipline? Who reviews a disputed result, and what underlying record can the worker see? The answers describe the actual role of a system in a particular workplace without assuming that one legal rule applies everywhere.

California’s boundary for automated discipline

California has signed protections aimed at some of these decisions. In a September 30 announcement, the governor’s office said the new legislation bars employers from relying only on an automated system for discipline or termination and addresses certain workplace surveillance practices. The signing is a confirmed step; it does not make every form of monitoring unlawful or give workers in other jurisdictions the same protections.

A score can still reach a manager before the worker understands the record behind it. When that score bears on a shift, a performance review or a disciplinary decision, the ability to see the underlying information and challenge an error becomes part of whether human review is meaningful.

Also read:

• HubSpot Cuts 660 Jobs—but Says AI Did Not Directly Replace Them