Синдикат Австралии предупредил: алгоритмы ИИ лишают работников права на отдых и выдают необоснованные предупреждения

20.08.26 In April 2026 the Shop, Distributive and Allied Employees’ Association (SDA), Australia’s retail and warehousing union, released For All of Us: Making AI Work for Working People, a report it commissioned from the John Curtin Research Centre. In response, UNI President and National Secretary-Treasurer of the SDA, Gerard Dwyer, spoke with UNI Global Union General Secretary Christy about the findings, about how artificial intelligence (AI) is changing retail work and the bigger picture of AI in society. “We are not robots”: AI on the shop floor For retail workers, AI is already here. It helps set the pace of their shifts and, in some stores, decides when they are disciplined. Gerard explains why the SDA commissioned the research and what it turned up. “With no human oversight, it went straight to head office and generated a warning.” Christy: Gerard, lets start with two easy questions. First, your union primarily represents workers in retail and warehousing – what inspired you to commission this research? And second, what problems do you see now, or foresee, for the workers the SDA represents? Gerard: We’re conscious of the advancing-technology component of retail. AI is coming, and we wanted to understand how it is going to disrupt our workforce. For example, we had been wrestling with these algorithmic rostering programs for a few years before people started talking about AI – but now it all makes sense. These are AI-driven algorithms for labour management, trying to match rostering to trading patterns and task completion, and we wanted to better understand what is going on. Another motivator was that when we develop a campaign we usually start with research. Good campaigns are built on solid research where data has been analysed professionally. So that was the motivation: the rostering algorithms causing problems we don’t fully understand, not fully understanding AI, what it is doing to our industry now and its capacity to do more in future. And in warehousing, we have wrestled with automation over the last decade, but AI will be another level on top of that, and we wanted to better understand its impact. Christy: On rostering – or scheduling, as it is called in North America – hasn’t this been automated for a long time? Are you finding the rostering algorithms are getting more sophisticated? Gerard: Much more sophisticated, much more problematic. We do have minimum shift entitlements of three hours, so that is somewhat of a bulwark against a one-hour shift for a particular peak, but it is still not good. We have problems accessing breaks, because the algorithm just doesn’t compute that you get a paid rest break. It can manage meal breaks, which are unpaid, because the computer sees you are not being paid and assumes you are off. But a paid break, where you are not supposed to do any work for 15 minutes, sends it into a spin. The computer doesn’t have the goods to handle it. The other thing is that we have no visibility of how the settings were arrived at. It appears to not account for our humanity and sets a pace that doesn’t understand we are not robots. You can’t pack and process cartons as fast on day four, in the fifth hour of your shift, as you could coming in fresh. We have also got a problem – not on rostering, but on another issue that shows why research is so important – with AI coming in above the checkouts to verify scan accuracy, and that goes straight up to head office. One of our store delegates was counselled for a scan rate of 99.8 per cent accuracy – that is about one error in 533 scans. The AI generated the counselling, and because there is no consultation with the union, it goes automatically up the chain to head office – not through people, just through little AI checkpoints along the way. It reaches head office, and this person with 30 years’ experience gets a written warning. If a human had overseen it at store level, I’m confident the department manager would have squashed it; if not, the store manager would have; if not, the organizer would have come in, and the warning would have been torn up. But with no human oversight, it went straight to head office and generated a warning. That is the sort of thing we are fearful of happening in other instances we haven’t yet thought of.