NICK DYRENFURTH, SEAN WHITWORTH & DOMINIC MEAGHER: Our AI bosses must respect human laws and dignity

At a customer fulfilment centre operated by one of Australia’s big retailers, a new automated loading system was installed last year.

A worker there told us the first notice came informally, from a colleague, a few months out.

When the system went live, staff in some roles had a few days’ warning, no manuals, and were expected to learn new screens on the clock. During a recent internet outage, the technology went down and there was no manual fallback. Staff raised IT tickets and waited.

At the same site, the overnight shift is now roughly 80 per cent casual and workers report that raising problems with the system can result in fewer hours.

Some full-time employees have been asked, voluntarily, to convert to casual contracts. A newer on-grid robotic picking system now handles somewhere between a tenth and a quarter of inventory picks.

Workers monitor it and record data so it can refine its grip for next time, training, in effect, the machine that may eventually do without them.

At another retailer, workers described AI-assisted rostering, automated hiring filters and centrally imposed labour benchmarks that local managers could barely override. Staff often only discovered a system using AI through conversations on the shop floor.

These aren’t imagined scenarios, but evidence acquired from the field for the John Curtin Research Centre’s new report, For All of Us: Making Artificial Intelligence Work for Working People.

They describe a pattern that has already settled into Australian retail, warehousing and logistics: consultation that happens informally if at all, a quiet drift toward casual contracts alongside each new technology, and opacity about where AI is being used.

The Albanese Government’s national AI plan is solid by international standards, committing to skills, sovereign capability and responsible public-sector adoption.

Crucially, it rejected the idea of a separate AI legal system. That is the right call. Australia already has laws governing discrimination, consumer protection, privacy, workplace safety and unfair dismissal. The task is ensuring those laws remain effective when AI becomes the instrument.

When AI allocates work, filters job applications or manages performance, it is exercising managerial power. Australia already has institutions designed to regulate power at work: the Fair Work system, collective bargaining and workplace law.

The task is not inventing an entirely parallel AI regime but ensuring technological change fits Australia’s long-standing settlement around fairness, safety, dignity and collective voice.

The NSW Parliament reached a version of the same conclusion in February, when it passed the Work Health and Safety Amendment (Digital Work Systems) Act. That law places responsibility for digital work systems on the employers that deploy them, not on the technology. It is a small precedent, but it points in the right direction.

Our report’s first recommendation is modest. A specialist expert panel on AI inside the Fair Work Commission, modelled on its existing panels, with the technical, legal and industrial capacity to set standards and resolve disputes involving workplace AI.

The second is more practical. For workplace AI systems that materially affect hiring, rostering, performance management, surveillance, discipline or termination, employers should be required to reach AI implementation agreements with their workforce.

These agreements would require structured consultation, transparency around how systems operate, guaranteed human review of significant automated decisions and ongoing workforce oversight.

Business opposition to consultation has already been overtaken by events. Microsoft has voluntarily signed a framework agreement with the ACTU covering workplace AI implementation, recognising that workforce trust and adoption are linked rather than opposed.

The productivity case for consultation is strong. International research increasingly suggests firms gain most from AI when workers trust the technology and help shape its deployment.

Yet surveys also show widespread fear of job displacement and declining institutional trust. Those conditions risk undermining the productivity gains AI promises before they arrive.

There is also a geopolitical dimension. The jurisdictions shaping AI governance now — from the European Union to major US states and India — are also positioning themselves inside the future AI supply chain. Drift is a position too, and the cost of it is not simply regulatory uncertainty but strategic irrelevance.

None of this is an argument against AI. Properly deployed, it can raise productivity, expand capability and improve work itself. Small businesses can now access tools once available only to large firms; regional GPs can draw on diagnostic support previously confined to major hospitals. Increasingly, the evidence suggests firms gain most when AI augments workers rather than simply attempting to replace them.

Our report also urges universal AI literacy, stronger sovereign capability, public procurement that rewards high-quality adoption rather than cost-cutting, and clear legal accountability that keeps humans — not algorithms — responsible for consequential decisions.

We have done this before. Faced with technological and economic upheaval in the 1980s, the Hawke government negotiated structural change through the Accords rather than imposing it from above.

Workers had a seat at the table while the economy transformed around them. Fairness became part of the design rather than an afterthought. Britain chose a harsher path under Margaret Thatcher. The working people who paid for Thatcher’s excesses kept paying for a long time and in time repaid the British political class with Brexit and the rise of Reform UK.

Today’s version of that unmanaged transition is being written elsewhere. But Elon Musk and the tech bros don’t build economic settlements or have Australia’s national interest at heart.

Our pro-worker approach to AI is a pro-enterprise strategy. Businesses require stable operating environments, skilled workforces and clear regulatory frameworks to invest with confidence. Getting the institutional settings right is essential not just for fairness, but for growth

The future of AI will not be determined by technology alone, but by the decisions made now about how it is governed, who benefits and whether it serves all Australians.

Dr Nick Dyrenfurth, Sean Whitworth and Dr Dominic Meagher are authors of a John Curtin Research Centre’s report into AI and employment