The next-generation Proteus changes the relationship between workers and automation fundamentally
Amazon on Thursday unveiled a warehouse robot that workers direct using plain conversational language — the same way they would ask a colleague to move something. No technical commands, no programming interface. The employee states the task; the robot determines the priority, the route, and the timing.
"You tell it what needs to be done. It figures out the priority, the route, the timing," said Scott Dresser, vice president of Amazon Robotics. "It becomes your assistant for material movement."
The next-generation Proteus was unveiled at Amazon's "Delivering the Future" event in Dartford, England, as part of a €10 billion investment in its European fulfilment network. Currently in lab pilots, it is planned for deployment in the first half of 2027. The existing Proteus operates at 25 US sites in dock areas only; the new model works across entire warehouse floors. Amazon also showcased STARK, a tote-handling system first piloted in Barcelona expanding to 15 European sites by 2027, and Vulcan, its first robot with a sense of touch, able to pick and stow approximately 75% of stored items at speeds comparable to frontline employees.
What actually changes when a robot understands plain language
When a robot is operated through specialist software, only trained personnel interact with it and the job boundary is clear. When a robot responds to plain language, every worker on the floor is a potential operator. The responsibility for supervising its decisions, correcting its errors, and managing its task prioritisation shifts to the person giving the instruction — and that is a new supervisory function that most current warehouse job descriptions do not capture and most current pay grades do not reflect.
One in three Australian organisations has already slowed entry-level hiring, with 60% expecting further reductions within three years, according to IDC research cited in HRD Australia's coverage of the Broken Ladder working paper from Warwick, LSE, and Oxford. Conversational robotics accelerates that pressure — but it also creates new roles that organisations deploying it without a redesigned job architecture will struggle to fill. Amazon's own data shows its Louisiana fulfilment centre, launched in late 2024, required 30% more employees in various roles because of advanced robotics. Those roles did not previously exist. Building the pathway to them — with a defined title, pay grade, and competency profile — before the robot arrives is the job HR needs to do now.
Amazon's engineers have also recently been found gaming the AI productivity leaderboards that HR designed — a pointed reminder that measurement architecture matters as much as deployment architecture. If you evaluate workers in robot-directed roles by task volume alone, you will produce the wrong behaviours. Accuracy, safety compliance, and exception handling need to be in the framework from day one.
The regulatory gap Australian HR leaders need to understand precisely
The federal government has taken a deliberate position: existing technology-neutral laws are sufficient to cover AI in the workplace, and no new AI-specific workplace legislation is coming. That position creates a specific set of obligations that Australian HR leaders need to map accurately — because the protections are narrower than many assume, and the compliance triggers that do exist are concrete and dated.
Enterprise agreement consultation clauses apply now. Many logistics sector enterprise agreements contain provisions requiring employer consultation before introducing major workplace change — and deploying a system that fundamentally changes how workers interact with their work environment, what skills they need, and how their performance will be evaluated qualifies as major change under most standard EA drafting. Workplace relations lawyer Shannon Chapman, speaking to HRD Australia's coverage of the John Curtin Research Centre AI report, noted that the absence of overarching AI legislation does not remove existing employer obligations — it makes understanding the intersection of those obligations more important, not less. If your EA contains consultation provisions and your operation is considering conversational robotics, the conversation with employee representatives should be happening in 2026, not when the equipment lands.
NSW WHS Digital Work Systems Act 2026 is live. The NSW Work Health and Safety Amendment (Digital Work Systems) Act 2026, passed 12 February, imposes specific WHS duties on PCBUs using AI or automation to allocate work. A system that receives and processes employee conversational instructions to direct physical tasks is, on a reasonable reading of the Act, within scope. Commencement is by proclamation, but the legislation is passed. Other states are watching. HR and safety teams in NSW logistics operations should be reviewing their WHS obligations against this Act's requirements now rather than at commencement.
Privacy Act automated decision-making obligations take effect 10 December 2026. Amended Privacy Act provisions will require businesses to disclose which types of decisions they make through automated means and what personal information they use. A conversational robotic system that captures employee voice instructions and processes them through an AI model to allocate and prioritise tasks is, on a reasonable reading, an automated decision-making system engaging personal information. HR teams evaluating logistics automation platforms in 2026 should be requesting Privacy Act compliance documentation from vendors before signing procurement contracts, not after deployment.
The contradiction at the heart of the announcement
The same day Amazon unveiled Proteus, three of its own engineers appeared before Seattle's Land Use and Sustainability Committee to oppose new large-scale AI data centre construction. The committee voted unanimously in favour of a one-year moratorium.
"It's been reported that this year, Amazon is spending $200 billion on capital, with most of it going to data centres and AI," said Patrick Schloesser, a software engineer at Amazon Web Services. "Meanwhile, the leaders at my company have laid off 30,000 corporate employees in the last eight months."
The 30,000 figure is Schloesser's characterisation. The structural pattern it describes — capital flowing to automation infrastructure while headcount falls — is one HRD Australia has documented in the technology sector throughout 2026. Amazon is pairing Thursday's announcement with a $1 billion Career Choice upskilling programme — part of a broader $2.5 billion Future Ready 2030 pledge — focused on cybersecurity, mechatronics, software development, and logistics.
"We couldn't find enough skilled people for the roles we need, so we made a decision: we're going to develop them ourselves," said John Boumphrey, Amazon UK Country Manager. The company plans to grow its European fulfilment workforce by 25,000 in the coming years.
Both things are true. The practical question for Australian HR leaders is which population they manage: the workers who will need to direct conversational robots in 2027, or the corporate headcount being reduced to fund the infrastructure that makes those robots possible. In most Australian logistics operations, the answer is the former. The job architecture, the enterprise agreement consultation, and the Privacy Act compliance work all flow from that.
Three things to act on before 2027
Redesign the job architecture before deployment, not after. The pathway from warehouse associate to robot fleet supervisor — title, pay grade, competency profile, and training programme — needs to exist before the Proteus lands, not as a reaction to it. Organisations that build this pathway retain workers through the transition and avoid the cost of redesigning roles under operational pressure.
Check your enterprise agreement consultation obligations now. If your EA contains a major change consultation clause, deploying a system that changes how work is allocated, supervised, and evaluated triggers it. Starting that process eighteen months before deployment is not premature — it is the timeline that allows the consultation to be genuine rather than performative.
Build Privacy Act compliance into your procurement process for any conversational AI robotics system evaluated in 2026. The 10 December 2026 automated decision-making disclosure obligations apply to systems that process personal information to allocate work. Request compliance documentation from vendors now. The window between procurement decision and the Privacy Act commencement date is shorter than it appears.
