Amazon's Atlas: Inside the Algorithm Designed to Predict Worker Unrest

Amazon has built a machine that purports to predict worker rebellion. According to leaked internal documents presented in legal proceedings by Unifor, Canada's largest private-sector union, the company deploys a proprietary AI forecasting tool called Atlas across warehouse operations in the US, Canada, and the UK to identify facilities at highest risk of unionization attempts or what Amazon internally terms "disruptive labour events."

Unifor alleges that Atlas ingests hundreds of data points at the warehouse level—worker communications, health and safety incident rates, local minimum wage, proximity of unionized facilities to non-unionized ones—to generate risk assessments. The union further claims, based on documents reviewed during British Columbia Labour Relations Board proceedings, that Atlas monitors external forums, specifically Reddit, to track sentiment "at the site level." Amazon has disputed these characterizations, stating that Atlas collects no personal data and functions only as an aggregated site-level sentiment tool to identify where facilities "need support."

The dispute emerged from certification proceedings after Unifor won union recognition at Amazon's YVR2 warehouse outside Vancouver in 2023. The British Columbia Labour Relations Board found that Amazon had engaged in unfair labour practices during the organizing campaign, including interrogation of workers about union sympathies and creation of an anti-union employee committee. However, the board's decision did not specifically adjudicate the Atlas allegations, which Unifor raised as evidence of systemic interference.

Unifor claims that two UK warehouse sites—BHX4 in Coventry and EMA4 in Minworth—were classified as "Tier 1" on Amazon's unionization risk scale, marking them for intensified monitoring. Amazon has not publicly confirmed the existence of this classification system or confirmed deployment of Atlas at these facilities.

Justin Gniposky, Unifor's organizing director, characterized the leaked documents as evidence of premeditated labour suppression: "Amazon's own documents pull back the curtain on what workers have been telling us for years. Behind the friendly PR, Amazon treats the fundamental right to unionize as a threat to be managed and eliminated."

Dr. Kathi Weeks, a professor of labour studies at Duke University who studies workplace algorithmic systems, offered a broader analytical frame: "Whether or not this specific tool functions exactly as described, the underlying logic is familiar. Employers have always tried to identify and discourage organizing before it gains momentum. What changes with algorithmic systems is the scale and the opacity. Workers can't see the model, can't challenge its assumptions, can't know if they've been flagged as 'high-risk' organizers."

Amazon's public response has focused on technical distinctions. A company spokesperson stated: "Atlas is a site-level forecasting tool that helps us identify operational areas that may benefit from additional support and resources. The characterization of this system as union-busting surveillance is inaccurate. We do not use personal data to track individual employees."

But labour law scholar Brishen Rogers of Temple University flagged a logical tension: "If an algorithm can identify which warehouse is most likely to unionize, it necessarily provides management with actionable intelligence about where and when workers may organize. Whether the company frames that intelligence as 'support' or suppression becomes a question of intent, not capability. The system's potential effect on organizing is real regardless of how Amazon describes its purpose."

The Atlas case sits within a broader context of tech-enabled worker monitoring that has become routine in Amazon facilities. Warehouses already deploy wristband scanners that track productivity, systems that monitor bathroom breaks, and algorithmic performance targets so granular that workers and labour researchers have described the experience as management by an opaque computational system. Atlas, if it functions as Unifor alleges, represents an extension of that logic: monitoring not current performance but future resistance.

The scale amplifies the implications. Amazon operates over 500 warehouses globally. If Atlas is deployed across a significant portion of those facilities, as Unifor's claims suggest, the company would possess an unprecedented labour-prediction infrastructure—every warehouse becoming a site of algorithmic risk assessment, every worker's communications and performance feeding a model whose output shapes management decisions about that worker's site.

Peter B. Olney, the retired organizing director of the International Longshore and Warehouse Union, assessed the allegations within historical labour context: "Union-busting is not new. What's new is the attempt to make it invisible and automated. A union organizer can challenge a manager who interrogates them. It's harder to challenge an algorithm whose logic is proprietary and whose decisions are rendered as neutral data."

What remains unresolved is the precise question of Atlas's actual deployment and effect. Unifor's allegations are supported by documents introduced in legal proceedings, but those documents have not been independently verified by journalists or subjected to cross-examination that would clarify contested claims about the tool's capabilities. Amazon's denial that the system targets individuals or operates as union-busting surveillance stands opposed to Unifor's interpretation of the same documents.

For workers at BHX4, EMA4, and the other sites mentioned in organizing campaigns, the uncertainty itself carries weight. Whether or not Atlas functions exactly as Unifor describes, Amazon has publicly acknowledged deploying algorithmic systems to assess site-level risk factors. The company's refusal to confirm or deny specific details about how those systems operate, or to explain what happens to facilities flagged as high-risk, leaves workers without clarity about the systems shaping their workplaces.

That opacity—whether intentional labour suppression or merely unfortunate corporate secrecy—remains the story's unsettled frontier.