Deloitte’s AI use created a blunder Down Under

Deloitte’s AI use created a blunder Down Under

Deloitte’s AI use created a blunder Down Under

The Australian government is getting a refund after a mistake-filled report



Anna Kim
ByDave Lozo
October 6, 2025

• less than 3 min read


You can’t make this up—but perhaps Deloitte’s AI could. The company is giving a partial refund to the Australian government after a $440,000 report was found to contain numerous AI hallucinations, the latest example of how consulting firms are grappling with the new tech.

What went wrong? Australia’s Department of Employment and Workplace Relations asked Deloitte to analyze the efficiency of its welfare system. But academics pointed out mistakes in the firm’s report that included citations to nonexistent studies, AFR reported. Deloitte admitted to using an LLM (Azure OpenAI GPT-4o) and later updated the report, but noted that the updates didn’t change its overall findings or recommendations.
A ‘human intelligence problem’

That’s how Australian Labor Senator Deborah O’Neill described the error. The phrase highlights the issues arising as AI takes over the grunt work formerly done by newbie consultants:The Wall Street Journal spoke to experts who predict consultancies will reap more benefits from AI work in four to five years, when the tech improves. And in May, Anthropic CEO Dario Amodei said that AI could wipe out half of entry-level jobs at white-collar companies within five years.
Harvard Business Review points out that this shift in hiring at professional services firms could leave companies wondering where partners, who do the less routine work now, will come from in the future.

Maybe more AI is the answer: Deloitte announced a massive deal yesterday to provide Anthropic’s Claude to more than 470,000 of its global employees over the next few months.—DL