When AI Overreaches: What the Deloitte Australia Case Teaches Us

Earlier this month, it was reported that Deloitte Australia will return part of a A$440,000 payment to the Australian government after issues were identified in a report prepared for the Department of Employment and Workplace Relations.

The report included non-existent academic references and a misattributed judicial quote, traced back to sections drafted using generative AI. Once the issue surfaced, Deloitte revised the document, disclosed its use of Microsoft’s Azure OpenAI service, and refunded part of the fee.

Officials confirmed that the report’s main conclusions remained intact, but the credibility of its supporting evidence was compromised.

This is not a scandal; it’s a lesson. A moment for reflection on how professionals integrate emerging technologies into serious advisory work.

The Real Issue: Judgment, Not Just Tools

Generative AI is a remarkable innovation. It can analyze, summarize, and produce text with impressive fluency, but it does not understand truth. It cannot distinguish between what sounds right and what is right.

That difference, between plausibility and accuracy, is where human expertise must prevail.

When consultants or strategists hand over too much thinking to an algorithm, they risk losing the foundation of their craft: judgment, experience, and ethical responsibility.

As we noted over a year ago in our article Beyond the Algorithm:

“Exploiting the full potential of large language models for strategy planning requires great expertise. You need to have the expertise to discern what is valuable from what is not. Asking an LLM to write a strategy for you will have limited value, and it could even be very risky. Expert human intervention is still crucial.”

That perspective feels even more relevant today. The Deloitte case is not about negligence, it shows how AI, when used without the right expertise, can quietly cross the line from helpful to harmful.

Using AI Wisely Requires Expertise

Technology can enhance how we think, create, and deliver, but it cannot replace understanding. Tools are only as capable as the people who guide them. Without intent, judgment, and verification, they can amplify mistakes as easily as they generate insights.

In advisory and strategy work, the challenge is knowing where expertise must lead and where technology can assist. That balance requires experience, not enthusiasm.

1. Expertise before automation. Understand the problem deeply enough to know what should, and should not, be automated.

2. Verify, then trust. Every fact, quote, and data point must be checked. Fluency is not accuracy.

3. Context and judgment. Experts bring nuance, ethics, and consequence, qualities no model can reproduce.

4. Clarity and transparency. Clients deserve to know how insights are produced and who is accountable. Trust grows when the process is open.

5. Guardrails that evolve. Organizations should embed responsible-use principles into their workflows: disclosure, validation, oversight, and accountability.


Expertise Is the Real Differentiator

As AI becomes more powerful, expertise -not automation- will define success.

The professionals and firms that thrive will be those who understand both the potential of technology and its limits. They will know when to accelerate with it, and when to stop and think.

That combination of human judgment and digital capability creates lasting value. It also protects what matters most: trust, reputation, and relationships.

A Balanced Perspective

Deloitte’s decision to acknowledge the issue and refund the client is commendable. The story is not about blame, it’s about balance.

AI is a powerful enabler, but it still requires direction, oversight, and restraint. Without expertise, it’s automation at scale. With expertise, it’s a true amplifier of human intelligence.

At ALG Group Global, we believe the future of consulting and strategy will belong to those who use AI wisely and responsibly, combining innovation with integrity, and automation with accountability.


Sources

Scroll to Top