NAB's AI job cut dispute: what NZ employers should document now

National Australia Bank chief executive Andrew Irvine told the ABC, in an interview published on 6 November, that the bank's 2,500 planned job cuts "had nothing really to do with AI at all." Irvine described the programme as old-fashioned simplification, with teams merged and leaders given wider spans of control. The Finance Sector Union had said weeks earlier that NAB executives refused to confirm whether AI was behind the cuts, and the dispute has since reached Australia's parliamentary AI inquiry. NZ employers should pay attention, because any restructure at a business rolling out AI will now draw the same question, and the answer carries legal weight here.

What happened

NAB announced on 20 October that it expected to cut up to 2,500 internal roles, about 6% of the total, in the 2027 financial year. Up to 1,500 of those would go before the end of 2026. The announcement listed rapid technological change among the pressures on the business and didn't mention AI. Swinburne University governance expert Helen Bird told the ABC the bank should come clean about AI's role.

NAB's public record on AI added to the pressure. Pete Steel, group executive for technology and AI, told a parliamentary committee on 9 October that the bank hadn't lost full-time equivalent staff to AI, while NAB has also told politicians it's using more AI, including to reduce manual work.

ACTU secretary Melissa Donnelly then took the NAB case to the Joint Select Committee on AI, arguing for a regulator with enforcement powers and saying employers too often bring in AI without meaningful consultation. The committee's terms of reference cover job displacement and gaps in existing law, and its final report is due by 30 November.

On 3 November the union reported concessions from NAB. Involuntary exits would move to 27 January 2027, consultation would run to 11 November, and the bank would supply an analysis of workload and staffing effects by 6 November. HRD Australia describes the January date more narrowly, as applying to staff in redeployment. NAB's analysis hasn't been published, so nobody outside the process knows whether it addresses AI. NAB's announcement doesn't mention New Zealand or its subsidiary BNZ, and any local effect is unstated.

Why the AI question is hard to answer

We can't tell from outside whether AI played a part in NAB's decisions. Irvine's statement and Steel's evidence may both be accurate. The union's suspicion is also understandable, given what the bank has said about technological change and its own AI use. Merging two teams is a standard simplification move, yet software that absorbs part of a team's workload can be what makes the merger possible. Few restructures will have a clean yes or no answer, so the answer needs evidence behind it.

In our view, NAB's exposure comes mostly from timing. The union's complaint was public by 21 October, and the question was still being put to Irvine after the A$6.14 billion full-year result on 5 November. An employer that can show its reasoning on the first day gives critics much less to work with.

What it means for NZ businesses

New Zealand votes today, and the AI settings could shift. Labour has promised to end the light-touch approach, and its policy includes an Office of AI, work with unions on safe AI use and standards aligned with Australia. The Council of Trade Unions wants upskilling, just transitions and regulation protecting workers' rights. If a Labour-led government forms, Australian recommendations could reach NZ employers faster. Businesses with staff in both countries should read the 30 November report whatever today's result.

The more immediate point sits in existing employment law, which applies whoever governs. Employment New Zealand lists improved technology as a genuine business reason for workplace change, so replacing tasks with AI can be a lawful basis for a restructure. The same guidance says employers should document their reasons and share them with potentially affected staff before deciding. Good faith duties require employers to be truthful and to avoid misleading anyone, though the dismissal process protections have an exception for employees on $200,000 or more who haven't opted back in.

Our reading is that the risk lies in the stated reason. An employer that presents a restructure as removing duplication, while the business case relies on an AI rollout, invites an argument that staff were misled and couldn't comment properly on the real proposal. The official guidance doesn't address AI directly, so take employment law advice on your own facts. Commercial sensitivity offers some protection, since employers can withhold confidential information for good reason, but that won't justify a wrong explanation.

A second risk arrives later. Employment New Zealand says a redundancy may not be genuine if a similar role is filled shortly afterwards. An AI tool that underdelivers can force exactly that rehire, which is a good reason to prove the tool's output before removing the role.

Practical next steps

• Record AI's role in every restructure paper. Note which tasks moved to software, which went through merged teams and what estimate supports each figure, and have a named executive sign the paper.
• Write the AI answer before the announcement. A short, factual paragraph gives managers the same wording when staff ask, and it should still hold up if the union asks for the working.
• Connect AI rollouts to workforce planning. When a pilot shows reduced workload, HR should know before headcount budgets are set, so the link is recorded early.
• Check your statements match. Compare what you've told staff, investors, customers and government about AI, since NAB's parliamentary evidence became part of its job cuts dispute.
• Run the workload analysis first. NAB agreed to supply one after pressure from the union, and an NZ employer can include one in the original proposal.
• Measure the tool before removing the role. Set a period of production use with error rates and rework counted, then decide.
• Decide what stays confidential and why. Record the reason for each withheld item, so you can explain it if consultation is challenged.

If AI is already changing how work gets done in your business, the governance and adoption work should sit alongside your HR process from the first pilot. Records written after a dispute starts will carry far less weight than records made at the time.