A guide that arrived in a bad week for trust
On Thursday 12 November 2026, Australia's federal government released a draft best practice guide on bringing AI into workplaces. The timing was awkward. That same week Qantas said it would move 850 domestic roles offshore through an outsourcing deal with Accenture. The Australian Services Union, which blamed AI for the cuts, warned that AI announcements mustn't become "a Trojan horse for widespread offshoring". The affected staff are due to leave from mid-2027.
Employment and Workplace Relations Minister Amanda Rishworth used the Qantas case to explain the guide on ABC Radio National. In her view those workers would have gained from direct consultation about how AI was being brought in. In a best practice scenario, she said, they'd also have been offered other roles. She also argued the guide could help keep hundreds of people in work.
For anyone building or selling AI tools to Australian and New Zealand employers, this is the important part. The guide doesn't regulate software. It describes how a well-run employer should roll AI out, and products that make that process easier will sit better with buyers who adopt it.
What's in the draft
The guide is organised in three phases: Plan, Enable and Embed. Under those phases sit eight principles. In our own summary, they ask employers to:
- work out where AI could genuinely add value
- pick concrete use cases and check whether the organisation is ready for them
- plan the change together with workers and involve them early
- rethink how tasks and roles fit together once AI is in the mix
- train people so they can use the tools well
- manage the effects on the workforce as roles shift
- put governance in place so it's clear who's accountable for AI
- deal with risks before they bite, including work health and safety
The draft includes case studies from employers. Brisbane Catholic Education reports that 71% of its educators now spend less time on admin thanks to AI. Commonwealth Bank says more than 32,000 staff have used AI-based tools for career planning, skills and learning. Blackmores says an AI-enabled process cut a manual data-entry job from three days to ten minutes. These are the employers' own figures. They haven't been independently audited.
The draft came out of months of roundtables run by the Department of Employment and Workplace Relations. Unions, employer peak bodies, academics and small businesses took part, along with large employers including Microsoft, Commonwealth Bank, Telstra, Kmart, Blackmores and Ramsay Health Care.
The risks it names
Reporting on the draft shows it's frank about how AI can make work worse. It warns that automating simple tasks can leave staff with denser, harder workloads. It flags monitoring features that end up being used for performance management. It notes that bringing in a tool without talking to staff can add a step to a process when the tool was meant to remove one. It also says entry-level work may be hit first, and that younger and older workers, First Nations workers, women and people with disability could be affected more than others. Higher pay, shorter or more flexible hours and more meaningful work appear in the draft as possible benefits. It doesn't promise them.
Rishworth has also pointed to a more ordinary problem she kept hearing about in consultations: companies buy AI tools and then nobody uses them, so the productivity gains never show up. Her answer is trust. In her words, employers "need their workforce to embrace AI".
What it doesn't do
The guide is voluntary. Rishworth said so directly: "This is voluntary." It creates no new duties and has no enforcement attached. Australian employers still have their existing Fair Work Act obligations to consult and to consider redeployment when they make major workplace changes or redundancies. Unfair dismissal law is technology neutral, so a worker can challenge a dismissal whether or not AI fed into the decision.
Critics say voluntary isn't enough. Independent Senator David Pocock welcomed the government's moves but said "we've lost years" in protecting Australians from high-risk AI, and he has pushed for mandatory reporting and copyright protections. The Greens' David Shoebridge has reportedly called for legal protections against surveillance and job losses. The guide also sits alongside a separate announcement made the same week: the government will bring in mandatory standards and a new regulator able to penalise technology companies that don't comply. Details of that regime weren't settled at the time of writing.
Why New Zealand builders should care
New Zealand already has voluntary guidance of its own. In July 2025 MBIE published the Responsible AI Guidance for Businesses together with the country's first AI Strategy. It suggests starting with a small project that solves a real problem, giving staff basic AI literacy, and choosing tools and providers that meet New Zealand's legal and ethical standards. It's lighter on worker consultation than the Australian draft.
New Zealand employment law fills some of that gap. Employment New Zealand's guidance on workplace change lists improved technology as a genuine business reason for restructuring. If an employer goes down that path, it must tell likely affected staff why, give them reasonable time to respond and suggest alternatives, and weigh their feedback before deciding. Staff are entitled to the information the employer holds about them, and the Employment Relations Act 2000 and Privacy Act both apply. So an AI rollout that changes jobs in New Zealand already triggers a consultation process. The Australian draft just spells out what doing that well looks like.
Many Auckland software firms sell across the Tasman. If large Australian buyers adopt this guide, it'll start appearing in procurement questions and pilot plans.
What builders can do with it
We think the guide works as a product checklist. Here's what we'd look at:
- Support staged pilots. Plan, Enable and Embed lines up with a small trial, a wider rollout, then steady use. Pricing and feature flags that allow a contained pilot make the first phase easier.
- Make consultation easy. Give employers plain-language explanations of what the tool does, what data it uses and which tasks it changes, written for staff as well as IT teams.
- Be careful with monitoring. If your product logs individual activity, make that visible, configurable and off by default where you can. The draft names monitoring as a hazard.
- Measure real use. Since unused tools are a known problem, adoption dashboards that show whether staff actually benefit help buyers keep the case going.
- Help with governance. Audit trails, role-based permissions and clear records of who approved which AI use make the governance principle easier to meet.
- Build in training. In-product guidance supports the capability principle and cuts the load on employers.
How to have a say
Consultation on the draft is open until 27 January 2027 through the Department of Employment and Workplace Relations consultation hub. New Zealand vendors with Australian customers can make submissions too. Sharing practical evidence about what makes AI adoption stick is a useful contribution while the guide is still a draft.
Sources
- First workplace artificial intelligence guide released (AAP via Neos Kosmos)
- AI productivity gains should be shared (Yahoo News)
- Draft best practice guide for AI adoption in the workplace released (The National Tribune, ministerial release)
- Draft AI workplace guide built with Microsoft, CBA, Telstra (OnNotice)
- Australia publishes draft workplace AI guide urging employers to consult workers (Superpower Daily)
- The basics of workplace changes (Employment New Zealand)
- New guidance to help you use AI responsibly (business.govt.nz)
Build with people working on the same things
Join the Auckland Society for AI: NZ$599 a year with a ForgeBox and the members' projects area. Our Auckland meetups are open to everyone.