The 113-page decision document that reshaped NZ Police’s information services never mentions artificial intelligence — not once — even though the restructure it authorised cut 41 file coordinator roles last month and will guide how the force’s back office absorbs automation for years. The document, released to RNZ on Tuesday and issued to staff in July, instead uses “automation” and “digital enablement” about a dozen times, and “transparency” almost 20.
Police told RNZ the same day: “We appreciate there has been some uncertainty about what the future of AI in the police will look like.” The gap between that sentence and the document’s vocabulary is where the story lives.
What the restructure did
RNZ’s reporting shows the year-long review significantly affected 72 of the 260 staff consulted, who work across 10 locations processing files vital to investigations and prosecutions. Police cut more than 40 positions and created about 20 in their Frontline File Management team — 41 of the cuts among file coordinators, the very roles the document describes as “critical” to maintaining quality and hitting deadlines. Seven staff were made redundant and 17 took voluntary redundancy.
The document’s own feedback section records what staff told leaders in roughly 800 submissions: that removing the coordinator layer “may create gaps that could affect accuracy, increase errors and slow turnaround times”, and that “uncertainty about how expertise will be recognised is contributing to lower trust in outcomes”.
There is a quiet irony in the unit’s name. The Government has repeatedly promised public sector job cuts would not touch the “front line” — but this restructure landed on the Frontline File Management team, including front-counter staff who are often the public’s first contact with Police.
The word that never appears
Police’s national manager for front-line file management and transcription services, Sarah Hutchings, told RNZ “artificial intelligence was not a factor of workforce reductions in this review”. Police’s July document says something different about the trajectory: “As these technologies become fully embedded, they will directly influence the size, structure, and skill mix of our workforce.”
Both statements can technically be true — no single role was cut because of an AI tool — but as B2B News puts it, the two claims sit awkwardly together: “the tasks went first. The roles followed.” Police has measured that effect itself, at least once. A BusinessDesk investigation cited by B2B News found Police’s $8.2 million Appian file-assessment platform cut per-case processing from 8–10 minutes to 1–3 minutes, saving 18,302 hours and roughly $997,000 a year — the kind of hard before-and-after data almost no other NZ agency has.
Meanwhile the unit staff feared for since 2024 — Transcription Services, which lived through two years of “panic” as police introduced AI transcription tools — was largely spared the cuts, but handed its own “modernisation programme”. A new Axon interview-transcription tool rolled out Monday, per RNZ’s companion report.
The wider pattern this fits
None of this happens in a vacuum. Documents released earlier this year show the Public Service Commission’s push to cut thousands of public sector jobs depends on “a much more aggressive/targeted approach to AI” — RNZ’s investigation of the 16 documents behind the 8,700-job reduction found Public Service Commissioner Sir Brian Roche describing the 55,000-headcount target as “a minimum so that as the transformation is implemented the full impact and value of AI is captured”. Police set up an AI innovation lab and adopted transcription, translation and investigative tools over the past year. The restructure vocabulary — efficiencies, enabling functions, FTE reductions — is the generic grammar of that programme.
Our read: the deliberate absence of the term “AI” from the document is the least transparent thing about it. Automation is the AI story at the workload level — file assessment, transcription, routing — and everyone in the consultation knew it. Naming the technology would have forced questions about which tools drive which savings, and police’s own history here is uncomfortable: the 2025 Whisper trial was pulled from English and te reo Māori transcription after officers breached safeguards, and internal documents attributed the ban to “the potential impact on specific roles” rather than technical risk. When an organisation pauses a tool to protect jobs, then restructures the adjacent team while its paperwork avoids the word, staff will read the pattern accurately no matter what the press line says.
The honest version costs nothing: measure before-and-after like the Appian rollout did, say which tools are coming, and concede that “digital enablement” shaping the workforce is the same claim as AI shaping the workforce, written in the passive voice. Transcription Services is already queued for its own modernisation round — the next document, like this one, will presumably find a softer word for it.
As of publication, Police had not responded to requests for comment beyond its statement to RNZ; the restructure document itself was released after RNZ inquiries rather than proactively.
