Wellington mayor Andrew Little said in a Newstalk ZB interview on Tuesday that “large chunks” of a $435,000 Deloitte report on the council’s staffing levels were “written by AI”. He didn’t specify which sections or how much of the report, and his office pointed questions from Local Democracy Reporting to a council paper detailing errors instead.
The report at issue is Future Fit Pōneke, released last November, which claimed Wellington City Council had 330 excess staff. That figure has been under sustained pressure since. Deloitte was found to have double-counted current employees alongside vacant roles, relied on three-year-old employment data sourced from the Taxpayers’ Union, counted every employee at their full-time salary regardless of hours worked, treated any role with “project” in the title as a duplicate job, and overstated staffing expenditure by $21.5 million. The council paper submitted last week concluded the assumptions behind the 330 full-time-equivalent recommendation were “inherently flawed”; a more defensible number would be 284, before accounting for the other assumptions.
Little was measured in one respect and blunt in another. He acknowledged the report offered useful insight — particularly its finding that the council’s contract management systems were a “disaster” — but described the rest as “frankly, just a waste of time”. He has previously called for a discount on the bill.
Deloitte’s response
Deloitte has not commented on the AI claim specifically. Asked for comment after Tuesday’s interview, the firm — which has previously stood by the report — referred to a statement from last week: “Deloitte uses a range of tools and technologies across its work, including AI-enabled tools where appropriate.” A spokesperson added that a human needed to sign off any analysis, findings and conclusions, and that the firm was “satisfied that the report’s conclusions reflect the work performed and the professional judgement applied throughout the engagement.”
That statement matters for how this should be read. Using AI-enabled tools is not, by itself, an error — and no one has alleged the report’s problems were caused by AI. The documented errors are data-handling failures: double-counting, stale data, salary assumptions. Little’s comment lands in a report that was already in trouble for reasons a person signed off on.
Why this story is bigger than one council
The reason this matters beyond Wellington is the accountability structure. When a consultancy delivers AI-assisted analysis, the client is buying the firm’s professional judgement, and the fee reflects that. If large parts of the document were machine-drafted, the question isn’t whether AI was used — it’s whether the review process matched the $435,000 price tag. A human signature on a document doesn’t tell you how much human scrutiny sat behind it.
There’s also a local irony worth naming. Little has been simultaneously attacking the report’s credibility and relying on its one finding he likes — that contract management is a “disaster”. That’s not hypocrisy so much as the normal messiness of using flawed evidence, but it shows how hard it is to unwind a consultant report once it has shaped a public conversation about “right-sizing”.
For New Zealand’s public sector, which is actively deploying AI tools across agencies, the takeaway is unglamorous: procurement contracts for AI-assisted work should specify disclosure of what was machine-generated and what verification was applied. We’ve looked before at how AI-compliance obligations are landing in New Zealand, and this episode is a case study in why the verification step — not the tool — is where liability actually sits. A $435,000 report with a $21.5 million arithmetic error would be a scandal either way. The AI detail just makes people ask harder questions about what they were paying for.
— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.
