Western Australia seeks repayment over flawed Deloitte merger report

By Elena Vasquez

The Western Australian government has sought expedited cost recovery for Deloitte’s $166,000 university-merger study after two cost scenarios carried discrepancies of $260 million and $311 million.

The study was meant to assess a potential merger among the University of Western Australia, Curtin university and Murdoch university. A WA government review will now test the report after cost errors and AI-hallucinated references emerged.

What did the merger report get wrong?

The state government paid Deloitte $166,000 for the feasibility study. The consultancy considered three options among the universities, assigning “low,” “mid” and “high” cost scenarios and showing the calculations in an appendix.

Under the “high” cost scenario for a merger between Curtin and UWA, the report put “total costs to day 1” at $514 million and “post day 1” costs over five years at $145 million.

Those figures add to $659 million. The report wrote $399 million, creating a $260 million discrepancy. Read another way, its stated total sat $145 million below the sum of its own day-one and five-year figures.

Under the “low” scenario for all three universities, day-one costs were $396 million and five-year post-day-one costs were $116 million. Those figures total $512 million, yet the report’s combined number was $823 million, a $311 million discrepancy.

That second total ran $311 million above the report’s own components. The two published scenarios therefore point in opposite directions: one understates the stated sum, while the other overstates it.

Prof Stephen Bartos, an expert in public sector governance at the University of Canberra, said the errors discredit the report.

“In my view, they’re so extensive that you probably need a new report, or at the very least someone to go through that material from the ground up and rework it manually without the use of AI,” he said.

“Any errors in a report like this, especially on something controversial like a university merger, discredits the report, and Deloitte shouldn’t have had these errors.”

Bartos said the large divergence between the figures was itself a cause for concern. “Unless the numbers add up, a government might not want to do it,” he said.

Where did AI enter the report?

Deloitte admitted in the report that it used generative AI to “support the collation of research, synthesis of information and editorial proofing.” The firm had earlier dismissed the errors as “minor.”

Pressure rose earlier in October after media reports revealed that the same report contained hallucinated AI references. The state opposition labeled the report “AI slop.”

Guardian Australia found that one reference to a World Health Organisation report carried a ChatGPT metadata tag on its URL, indicating that the link came directly from the chatbot.

The report also repeatedly referred to Curtin university as the “Curtin University of Technology.” The institution has not used that name since 2010. State parliament changed it under an act in 2017.

Deloitte’s admission placed generative AI across three stages of report production rather than one isolated task. The company said the technology supported research gathering, information synthesis and editorial checking.

Why is WA seeking the money back?

The Department of Energy and Economic Diversification wrote to Deloitte to expedite the cost-recovery process, a WA government spokesperson told Guardian Australia.

“We are extremely disappointed these errors are being found now,” the spokesperson said.

The spokesperson said: “We are advised they do not affect the modelling in the report or its findings, but any errors will be corrected as part of a review by Deloitte which will re-issue the report.”

“All available data, along with new information, will be interrogated by the university reform business case,” the spokesperson said.

Tony Buti, the WA tertiary education minister, first described the errors as “typos.” A day later, he said he was angry and disappointed with Deloitte.

Deloitte did not respond to questions from Guardian Australia.

Can the merger still happen?

The WA government has argued that the merger is a “once-in-a-generation” opportunity. Buti told the ABC on Tuesday: “I want to commit to it if the business case stacks up, and then it’s a government decision.”

Support does not extend across all three institutions. Murdoch university did not back the proposal.

Prof Andrew Deeks, Murdoch university’s vice-chancellor, criticised the report in an op-ed on the university’s website. He wrote: “all that a university merger will do for WA is cost a lot of time, money and energy, while reducing choices for students and alienating alumni.

“That the Western Australian government continues to be taken in by this nonsense is extraordinary,” he wrote. “Another fail for Deloitte.”

Deeks has rejected the merger itself. Buti has kept the decision open, saying it belongs to the government if the business case holds.

What should government consulting contracts require?

Prof Christian Downie, from the school of regulation and global governance at the Australian National University, has previously examined submissions to government containing AI hallucinations. He said human errors happen and mistakes in reports can be expected.

Downie said Deloitte’s broad description of its AI use offered too little transparency. “I think we need much more detail about how [AI is] being used and for what parts of the report writing process,” he said.

“If they’re going to use it for data collection and if there’s concerns about the data, then that needs to be detailed in the report,” Downie said.

Without that detail, Downie said, it was hard to have confidence in the report. As AI use increases, he said, governments need to reassess their contracts with consultancy firms.

“I think all levels of government, both state and federal, need to be recalibrating their contractual relationship arrangements when they outsource work to consulting firms to ensure that the proper incentives and penalties are in place in those contracts,” he said.

Downie’s prescription is practical: specify how AI was used, which data it touched and what penalties apply when source checks or arithmetic fail. Deloitte’s admitted use covered research collation, synthesis and proofing, three separate points where an error could enter a public report.

The immediate test is whether Deloitte can reissue a report whose totals reconcile and whose AI-assisted work can be traced in detail. If it cannot, the minister’s condition — that the business case “stacks up” — will not have been met.