Australia convenes employers to draft the country's first AI-at-work guidance

You have probably sat in a room where someone promised that AI will "augment, not replace" your team, then watched the conversation stall on what augmentation actually looks like when you hand a claims processor or a call-centre analyst a generative model. Last week seven of Australia's largest employers – Commonwealth Bank, Microsoft, AGL, Telstra, Kmart, Blackmores Group and Ramsay Health Care – gathered in Canberra to answer that question in front of Employment and Workplace Relations Minister Amanda Rishworth. The roundtable was the first of five government consultations designed to produce Australia's first tripartite Best Practice Guidance for AI in the workplace, a document that will carry the imprimatur of unions, employer peak bodies and the Fair Work Commission.

The timing is deliberate. Jobs and Skills Australia reported in 2025 that AI tools will more likely augment work than replace workers, but augmentation is a category error if you do not specify who owns the productivity dividend, who retrains the worker whose task shrinks and who decides which roles disappear when a workflow compresses. Rishworth told ABC Radio National Breakfast that the guidance will answer a simpler question: "What does best practice look like when adopting AI in workplaces?" The phrasing is careful. Best practice, not minimum compliance. Adoption, not deployment. The minister is signalling that the document will not be a checklist for legal teams but a playbook for line managers who need to explain to a rostered nurse or a branch teller why the system now writes half their notes.

What the roundtable covered

The agenda centred on three questions. First, what are the foundations of best practice AI adoption that share benefits between workers and employers? Second, what strategies enable effective and inclusive change management when a tool arrives that rewrites job descriptions faster than HR can update them? Third, how do you build workers' trust and confidence when the technology is opaque and the vendor roadmap changes every quarter?

The participants were chosen because they have already introduced AI into workplaces that span retail floors, hospital wards, call centres and energy grids. Commonwealth Bank has embedded machine-learning models into credit decisioning and fraud detection for years. Microsoft sells the tools that other employers buy. Telstra and AGL operate networks where predictive maintenance and demand forecasting are table stakes. Kmart runs supply chains that hinge on inventory algorithms. Blackmores and Ramsay Health Care employ knowledge workers – pharmacists, clinicians, compliance officers – whose roles are adjacent to the tasks that large language models now perform at speed.

The roundtable did not produce a draft document. It collected case studies. Rishworth said the lessons shared will "help inform Australia's first tripartite Best Practice AI Guidance and make sure it reflects the realities of modern workplaces." The phrase "tripartite" is the load-bearing word. The guidance will not be a government white paper or an industry code. It will be a negotiated text that unions, employer groups and the state agree to endorse, which means it will carry moral weight in Fair Work disputes even if it never becomes statute.

Why this matters for L&D and innovation leaders

If you lead learning, digital transformation or innovation inside an Australian enterprise, this process will shape the questions your executive asks when you propose an AI pilot. The guidance will not tell you which vendor to choose or which use case to prioritise, but it will establish the threshold for what counts as inclusive change management and what counts as consultation theatre. When a union delegate or a works council representative asks whether your rollout meets best practice, they will point to a document that your CEO's industry body helped write.

The practical stake is this: start now to document how your AI pilots share productivity gains, how you retrain workers whose tasks compress and how you explain model behaviour to people who do not read Python. The guidance will codify what large employers already do when they do it well, which means the gap between your current practice and the published standard will be visible to your board, your regulators and your workforce. Rishworth's framing – "if business and workers are going to have the confidence to embrace the benefits that come with AI, we need to look at what does best practice in that adoption mean" – is an invitation to get ahead of the guidance rather than retrofit compliance after it lands.

The four remaining consultations with unions, small businesses, employer peak bodies and academics will run through the rest of the year. If you operate in Australia and you are deploying AI in customer service, operations, compliance or knowledge work, the case studies you build now will either vindicate your approach or expose the distance between your change-management rhetoric and the experience of the people who use the tools every day.

Sources:

• HCA Mag, 21 September 2026