Amazon's Atlas tool scores warehouses for union risk, and the staff survey is one of its inputs

openDemocracy reported this week that Amazon uses an in-house AI forecasting tool called Atlas to predict how likely staff at its warehouses in the US, Canada and the UK are to form a union, stage a sit-in or take part in what the company calls "disruptive labour events". The internal documents came out in a legal challenge brought by the Canadian union Unifor and were shared with the US non-profit publication Truthout. openDemocracy traced the UK sites named in them. The documents say Atlas gathers "hundreds of data points", including Reddit posts and staff responses to internal feedback surveys, and gives each warehouse a score between 100 and 900. A lower score means workers are more likely to organise. For leadership, the document states, higher scores are "better".

What the documents describe

Warehouses fall into five tiers. Tier 5 means "no immediate perceived risk". Tier 1 means "future activity is imminent". According to the document, the scale can "inform staffing decisions, resource planning, standard actions, and ownership to mitigate risks to Amazon labour structure".

Sixteen warehouses are Tier 1. Two are in the UK: EMA4, a £500m robotic-fulfilment centre in Sutton Coldfield, and BHX4 near Coventry. Coventry is where around 2,600 workers took part in an official ballot for recognition of the GMB union in July 2024. GMB lost by 28 votes.

Amazon's position is that Atlas covers "site/building level activity only". It says it does not track individual employees' opinions, affiliations or union membership, and that Atlas helps it "understand where associates need more support". The company told openDemocracy that a framework called Atlas is not currently used in the UK. When openDemocracy published, Amazon had not said whether it uses a similar system under a different name.

The feedback channel is also a data feed

The detail that matters for anyone in a job is the input list. The internal feedback survey is the channel an employer offers so staff can raise problems. Amazon's spokesperson said the company uses that feedback "to identify where we can improve the day-to-day experience of our employees". The documents show the same responses also feed a model that forecasts collective action.

One dataset can serve two purposes. Whoever runs the model decides which one applies.

Garfield Hylton, 62, works at the Coventry site. He described a power imbalance: the employer holds the aggregated picture across hundreds of signals, while workers have only their own experience to go on. "It's not right in a democracy… the employer should not be interfering," he told openDemocracy.

Regulators have ruled against Amazon on this before. In 2022 France's data watchdog CNIL fined Amazon €32m for breaching GDPR, because its tracking of workers was precise enough to trigger an alert after one to 10 minutes of inactivity. Separately, the British Columbia Labour Board ruled that Amazon had violated multiple local labour laws. A UK government spokesperson told openDemocracy that monitoring technology should be used "transparently and in ways that are fair, accountable and reliable".

This pattern reaches well beyond warehouses. Any employer with engagement surveys, an internal message board and an analytics budget can build a score. AI does the routine work of pulling the signals together. People still decide what the score is for, and that decision is where you have a say.

Your Next Move

1. Ask what is measured. Ask HR in writing which tools process staff survey responses and internal-platform posts, at what level they are aggregated, and what decisions they feed. If you work in the UK or EU, a subject access request is the formal route to the data held on you.

2. Write feedback as evidence. Treat every internal channel as recorded and modelled. Keep each point specific and dated, and tie it to an outcome the business already tracks. A record written that way serves you whichever model reads it.

3. If you manage people, question the score before you act on it. Find out which inputs produced it and what the tier is meant to trigger. Accountability for a staffing decision stays with the manager who makes it, whatever the model said. For how employers already measure staff with AI tools, and what the law gives you, see what to do before your employer grades your AI use.