New Zealand Police cut roughly two dozen jobs in its file management unit last month, part of a restructure that disestablished, created or shifted more than 60 positions altogether. The detail that caught career watchers off guard: transcription services — the unit whose staff had voiced job fears since police began rolling out AI transcription tools in 2024 — was largely spared, as RNZ reported.
The story matters beyond one agency. It is a case study in how AI reshapes a workforce in practice, which is rarely the same story as the one predicted in advance.
What the restructure actually did
Police released a 100-page-plus restructure document to RNZ, dated July, with job changes kicking in during October. Front-line file management and transcription services national manager Sarah Hutchings told RNZ the restructure aimed for a more nationally co-ordinated, flexible service that can “leverage digital and automation efficiencies”.
Police told RNZ they aimed for clearer accountability and efficiency, and that no jobs were cut by AI itself — the changes came from restructuring how units are organised around new tooling. The B2B News writeup of the restructure carried the same framing: police planning a smaller, AI-shaped workforce rather than announcing AI replacements role by role.
The unit that feared AI most was not the one cut
Transcription services staff have talked about “panic” over job security since police began introducing AI tools to speed up investigations two years ago. When the team heard police in the regions were using Whisper, OpenAI’s transcription model, in mid-2024, they voiced job fears and contacted the Police Association, according to reports and emails released to RNZ.
The manager of intercept and technology operations acknowledged the anxiety in March 2025: “There was a recognition that automating English transcription could affect positions traditionally responsible for this work,” Stuart Mills said in a report.
Yet transcription has been largely spared. Front-line transcription services national manager Sarah Hutchings told RNZ that people were still needed to transcribe accurate, quality statements — such as for court. Accuracy requirements, not enthusiasm for AI, decided which roles remained. Police also rolled out a new interview-transcription tool from an American tech firm on Monday, and this year added a second translation-transcription tool made by Meta, alongside European law enforcement AI tools for child exploitation investigations.
The union’s concern
The Police Association, which has about 280 members in the restructured unit, told RNZ that AI came up in its talks with police about the changes. The PSA union said it had not been able to establish instances of consultation with police about the use of Whisper or any other AI transcription tools. Police have said they take a considered approach to adopting new tech, weighing operational need, risk, legal requirements and policy considerations, and acknowledged “there may be some anxiety from Transcription Services staff because these tools have been approved for use”.
For workers, the sequence is worth noting: the fear arrived two years before the cuts, and the cuts did not land where the fear did.
A wider labour market reading the number against
The restructure lands in a soft labour market. Stats NZ’s September-quarter household labour force survey, released this week, put the seasonally adjusted unemployment rate at 5.7 percent, up from 5.6 percent in June, with the underutilisation rate at 13.8 percent and 174,000 people officially unemployed, according to the Stats NZ information release. Against that backdrop, a two-dozen-role cut in one back-office unit is a small number with a loud echo: back-office roles are where public agencies are already testing how far AI-shaped restructuring can go.
It also fits a pattern Singularity.Kiwi has tracked across the public sector. The government’s plan to replace thousands of public servants with AI has already drawn analysis about its hidden costs in our NZ public sector AI coverage, and the wider question of whether AI creates or removes openings is running on two separate margins, as we covered in our piece on economists splitting AI’s jobs effect.
What this means for back-office careers
Three practical signals come out of this single restructure.
First, AI-adjacent cuts follow accountability lines, not automation lines. File management — routing, checking and closing case files — was restructured hard. Court-grade transcription, where a wrong word has legal consequences, held its headcount. Where output can be verified cheaply, AI sits beside the work; where verification is expensive, humans stay.
Second, the anxiety arrives years ahead of the restructuring. Transcription staff were warning about Whisper in 2024; no transcription roles were cut in 2026. For anyone managing a career through an AI rollout, the tool arriving in your workflow is not the same event as your role disappearing — though the two can be connected over longer horizons than a single restructure reveals.
Third, agencies reorganise once, not role by role. Police combined disestablishments, new roles and shifted positions in a single 60-plus-position redesign. Workers in any large organisation about to “leverage digital and automation efficiencies” are more likely to see one big restructure than a slow drip of individually automated tasks.
FAQ
Did NZ Police say AI replaced the jobs they cut? No. Police told RNZ the job changes came from a restructure aimed at a nationally co-ordinated, flexible service with digital and automation efficiencies — not from AI replacing specific roles one by one.
Which unit was cut, and which was spared? Roughly two dozen jobs went in the file management unit in October, in a restructure touching more than 60 positions. Transcription services — the unit that had feared AI transcription tools like Whisper since 2024 — was largely spared, with police saying people are still needed for accurate, court-grade statements.
How many people were unemployed in NZ as of the latest data? Stats NZ’s September-quarter survey, released 4 November 2026, counted 174,000 unemployed people, a 5.7 percent unemployment rate, and a 13.8 percent underutilisation rate.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
