Amazon uses an in-house AI forecasting tool called Atlas to predict how likely its warehouse workers are to unionize, stage sit-ins or stage other "disruptive labour events," according to internal documents made public as part of a court battle between the company and a Canadian union. The investigation, published by openDemocracy on October 16, suggests the tool assigns each warehouse a risk score of 100 to 900 — and that leadership is told higher scores are "better."
What the Documents Say For more context on this story, see our ongoing breaking AI news.
The documents, which openDemocracy reports appear to be training resources for Amazon managers in Canada, contain an appendix stating that Atlas amasses "hundreds of data points" about the workforce. Those inputs reportedly include posts on Reddit forums, employees' responses to internal feedback surveys, general workplace information, and any previous industrial action at the site.
From that data, Atlas generates a score between 100 and 900 for each warehouse's likelihood of union activity. The lower the score, the more likely the system believes workers — whom Amazon calls "associates" — are to unionize or take other actions that could affect the company's productivity or profitability. The document states that for Amazon's leadership, higher scores are "better," which openDemocracy notes makes clear the company treats unionization as a risk to be reduced.
Warehouses are categorized into five tiers based on their Atlas score. Tier 5 represents "no immediate perceived risk," while Tier 1 means "future activity is imminent." The documents suggest the scale is used to "inform staffing decisions, resource planning, standard actions, and ownership to mitigate risks to Amazon labour structure."
Inside a Scored Warehouse
The report grounds the story at Amazon's Coventry warehouse in the UK, where 62-year-old worker Garfield Hylton told openDemocracy he has long known that Amazon monitors and logs every minute of his working day — every product scanned, every second between scans, every moment his scanner sits briefly inactive — all of it feeding productivity targets and predictions.
What Hylton said he did not know is that Amazon appears to also monitor potential "dissent" at the Coventry site, using AI to surveil posts on social media and the company's internal staff bulletin board.
Amazon's Response
An Amazon spokesperson refuted the characterization of Atlas as "individual employee monitoring or surveillance," telling openDemocracy the system operates at the site and building level rather than tracking individuals. The company did not dispute the existence of the forecasting tool itself, according to the report.
Michael Newman, a partner at London-based law firm Leigh Day, told openDemocracy that the scoring system "makes it clear that they think the union is bad," and that it tells "a very different story" from Amazon's previous assertions that it does not seek to prevent workers from joining a union.
A Broader Reckoning Over Algorithmic Management
The revelations land amid intensifying scrutiny of how large employers use algorithmic systems to manage and monitor workforces. Amazon's warehouses have long been among the most heavily instrumented workplaces in the world, with productivity telemetry down to the second. What is new in the Atlas documents is the apparent fusion of that operational telemetry with external signals — social media posts, survey responses — into a forward-looking predictive risk model aimed at organizing activity specifically.
Labor advocates have argued that predictive union-risk scoring raises distinct legal questions from ordinary productivity monitoring, potentially touching worker organizing protections in the US, UK and Canada, the three regions the documents say Atlas covers. The court battle that surfaced the documents remains ongoing, and it is not yet clear what further material may enter the public record.
How Predictive Labor Scoring Works — and Why It's Controversial
Atlas belongs to a category of workplace systems sometimes called "labor forecasting": models that treat organizing activity as one more operational variable to optimize, alongside demand curves and staffing rosters. The documents describe a feedback loop familiar from other enterprise AI deployments — data collection, risk scoring, tiered response — but the "response" here involves decisions about staffing and resources at sites where workers are attempting to exercise organizing rights.
Critics of such systems argue that scores like Atlas's can create self-fulfilling dynamics. A warehouse flagged as high-risk may draw closer management attention, adjusted staffing, or "standard actions" whose effects fall on the entire workforce, regardless of whether any individual employee has engaged in organizing. Because the scoring criteria are not public, workers at a given site generally have no way to know whether their warehouse carries a Tier 1 rating or a Tier 5 one — or what data about them contributed to it.
The openDemocracy report does not allege that Atlas targets named individuals; the dispute centers on site-level scoring and what Amazon does with it. But the documents' description of Reddit monitoring and internal survey analysis has already drawn criticism from labor organizers online, where the story circulated widely after being discussed on Hacker News.
For the AI industry, the case is another data point in a debate that has moved from hypothetical to concrete: employers are no longer asking whether AI can predict workforce organizing — the documents suggest at least one of the world's largest companies built a system to do exactly that. Regulators in the UK and EU, already drafting rules around algorithmic management and workplace AI, now have a named, documented example to work with.
_This article is based on openDemocracy's investigation published October 16, 2026, by Evie Breese._
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
