Amazon warehouse workers are tracked from the moment they badge in to the moment they badge out. Their scan rates are monitored in real time. Their idle seconds are counted. Their bathroom breaks are logged, measured, and in documented cases, used as the basis for disciplinary action and termination.
Amazon calls this “time off task” monitoring. Workers call it what it is: surveillance so granular that it treats human beings as components in a logistics machine, to be optimised or replaced when they fall below output targets.
The System in Detail
Amazon’s fulfillment centre workers carry handheld scanners that record every item they pick, pack, or stow. The scanners stop recording when a worker puts them down. Every gap in the scan data is logged as “time off task.” The system has no way to distinguish between a worker taking a medically necessary bathroom break and a worker standing around doing nothing, so it treats both identically.
Internal Amazon documents obtained by journalists at The Verge and the New York Times showed that workers at some facilities were expected to keep time-off-task below 18 minutes for a ten-hour shift. Workers who consistently exceeded this threshold received automated warnings, then termination notices, in some cases without a human manager reviewing the underlying data.
Amazon has confirmed the existence of time-off-task monitoring while disputing characterisations of how it is used in termination decisions. The workers who received those automated termination notices tell a different story.
The AI Layer
What has changed in recent years is not the existence of monitoring but the sophistication and automation of the response to it. Early warehouse monitoring required a manager to review data and make a decision. Current systems use machine learning to identify workers whose productivity patterns deviate from the expected distribution and generate automated interventions without human review.
This is the direction all workplace monitoring is heading. Microsoft’s AI tools are being used to monitor office worker productivity in ways that are functionally similar to what Amazon does in warehouses, just with spreadsheets instead of scanners. The principle is identical: generate a continuous data trail of worker output, use algorithms to flag deviations, automate the consequences.
The Legal Situation
In the United States, employer monitoring of workers is broadly legal. There is no federal law limiting what data an employer can collect on employees during work hours, how it can be used, or whether workers must be informed of the specific metrics being tracked. Some states have passed limited protections, but the baseline federal position is that your employer owns your time at work and can measure it however they choose.
The European Union has taken a different approach. The GDPR and the EU AI Act impose transparency and proportionality requirements on automated systems used in employment decisions. An automated termination system of the kind Amazon uses would face significant legal scrutiny in most EU member states. It faces essentially none in the US.
What This Is Really About
Amazon’s monitoring systems are not primarily about catching bad workers. They are about removing the human element from the management of labour costs. A human manager deciding whether to discipline a worker who took too many bathroom breaks has to sit with the discomfort of that decision. An algorithm does not. The automation of discipline makes it easier, faster, and cheaper to treat workers as interchangeable units rather than people.
The workers building the AI systems that enable this monitoring are paid poverty wages to do it. The companies deploying those systems are valued in the trillions. That gap is not an accident.