From delivery routes to office dashboards, workplace surveillance is turning everyday activity into performance judgments. Research suggests the bigger danger is what happens after a score is calculated.
A nurse can speak calmly to a distressed patient, but an automated sentiment score may suggest she sounds insufficiently warm. A bus driver encounters traffic, yet software interprets the delay as poor performance. Neither example necessarily reveals how well someone did their job.
These tensions anchor an October 8, 2026 report, Bossware Comes to California, produced by the UC Berkeley Labor Center, UCLA Labor Center and partner organizations. Drawing on 25 in-depth worker interviews and existing research, it describes employers using digital systems to observe workers, direct tasks, evaluate output and influence discipline.
The bigger issue for software builders is how a measurement becomes an employment decision.
Surveillance Is Becoming Management Infrastructure
“Bossware” includes GPS tracking, biometric clocks, productivity dashboards, algorithmic scheduling, cameras and systems that analyze calls. Some tools use AI; others follow simpler rules. Their shared feature is converting workplace behavior into data managers can act upon.
A 2025 OECD employer survey of more than 6,000 middle managers across six countries found that 90% of surveyed US managers said their firms used at least one algorithmic management tool. The comparable average was 79% across France, Germany, Italy and Spain, and 40% in Japan.
Those percentages include software that instructs, monitors or evaluates employees; they do not mean every worker is continuously surveilled.
A separate European Commission survey of 70,316 workers across the EU found 37% had their working hours digitally monitored, 24% experienced automated allocation of working time, and 13% had performance automatically assessed and rewarded.
A Productivity Score Is Not Productivity
The technical problem is measurement validity. Software can count mouse movements, calls handled or packages scanned while overlooking judgment, collaboration, care and circumstances beyond someone’s control.
The California report documents workers disputing package-scanning errors, driving infractions and alerts they never received. It also describes delivery workers pressured to skip bathroom breaks and freight drivers dismissed following algorithmic rankings.
In the UCLA Labor Center’s account, a telephone triage nurse identified as Maya explains that staying “calm and neutral” during a serious medical situation can be penalized by low sentiment scores.
The problem is not necessarily faulty arithmetic. It is assuming a measurable signal adequately represents a complex human outcome. When rankings determine shifts or discipline, a disputed proxy can become a consequential verdict.
Once scores feed scheduling software, errors can compound: a lower rating might produce fewer preferred shifts, reducing earnings and future opportunities. Improving the accuracy of one metric does not automatically fix the downstream decisions built around it.
What Constant Monitoring Does to Workers
Surveillance may also change behavior before a manager reviews any dashboard.
A 2026 open-access German study by Sophie-Charlotte Meyer, Charlotte K. Schröder and Anja-Kristin Abendroth found automated storage of work data and perceived constant surveillance were negatively associated with job satisfaction. Perceived surveillance showed the stronger association, while job autonomy was particularly important.
These survey associations do not establish that monitoring caused dissatisfaction.
Another 2026 study in the Journal of Business and Psychology, led by Tara S. Behrend, analyzed a nationally representative sample of 2,515 US workers. Precarious workers were more likely to face intensive monitoring and report negative psychological reactions. More secure workers were likelier to see potential safety or productivity benefits.
The same technology may feel useful to someone who can challenge a score, and threatening to someone whose livelihood depends on an unexplained ranking.
Yet findings are not uniformly negative. A study published by Cambridge University Press, by Monique Delfim Andrade and colleagues, surveyed 382 hybrid and remote workers and found no statistically significant overall difference in psychological safety between monitored and unmonitored groups. Its analysis suggested implementation details may matter.
The Evidence Does Not Support a Simple Ban
Monitoring can have legitimate purposes. The International Labour Organization’s 2025 safety report notes that smart systems can reduce hazardous exposure and prevent injuries, while warning that algorithmic management introduces new occupational risks.
Even managers express reservations. In the OECD survey, nearly two-thirds of managers using these tools reported concerns. Unclear accountability for incorrect decisions led the list at 28%, followed by difficulty understanding algorithmic logic and insufficient protection of worker health, both at 27%.
The distinction is between monitoring that provides useful safety information and systems that automate punishment without context or appeal.
The Next Software Design Problem Is Accountability
For vendors, employee monitoring cannot remain merely a dashboard feature. Responsible systems should explain what is collected, distinguish activity signals from verified outcomes, enable corrections and preserve human review before consequential action.
That means auditing false alerts, testing across different roles, limiting data retention and giving employees access to records used in assessments. These are proposed safeguards, not claims that existing products universally provide them.
The California researchers argue that technology’s effects depend on who controls it and whether workers have enforceable rights. For founders, that reframes the product challenge: trustworthy management software must serve the people being measured, not only the managers buying it.
The question is no longer whether software can watch work. It is whether anyone can challenge the story that software tells about a worker.
