Somewhere between drafting, reviewing and publishing, the process failed to trigger the level of challenge the report required.
That isn't primarily an AI problem.
It's a behaviour problem.
And I suspect versions of it are happening quietly inside organisations every day.
We've always used "looks right" as a rough proxy for "is right."
It was never a perfect test. But it was usually a useful one.
Producing polished, well-structured, authoritative work took time and effort. And while effort never guaranteed quality, it often signalled that someone had wrestled with the thinking.
AI has broken that relationship.
Today, a first draft can look remarkably close to a final one. A confident recommendation can appear just as persuasive whether it's been rigorously challenged or simply generated in seconds. A fabricated reference sits in exactly the same font as a genuine one.
The polish is now free.
Our instincts, however, still belong to a world where polish usually meant thought.
AI has broken one of the rules organisations have quietly relied on for decades: if it looks right, it probably is.
Go back to the Deloitte story with that in mind.
The real question isn't why the AI hallucinated.
It's why the document never triggered the behaviours that might have caught it.
Pause.
Challenge.
Curiosity.
Ownership.
Everything about the report signalled that the thinking had already been done.
So nobody behaved as though it hadn't.
That's the shift I think matters.
Not whether organisations are using AI. Almost everyone is.
Not which models they've chosen. Those will change every few months anyway.
The more interesting question is:
What AI-related behaviours are strengthening organisational performance... and which are quietly undermining it?
I've seen the opposite too.
Teams using AI to generate possibilities they then challenge, refine and build on together.
Leaders using it to expose assumptions before making decisions.
People using it to get past the blank page, then doing the thinking that really matters.
AI removes the blank page. The question is whether it also removes the doubt.
Other behaviours are far less obvious.
Alternative ideas never really emerge because the AI's answer feels good enough.
People increasingly work one-to-one with AI instead of bringing their thinking into the open, so challenge gradually disappears from the process.
Nothing feels wrong.
In fact, everything feels faster.
Until one day the organisation starts noticing that originality has declined, decisions feel strangely similar, or people have become less confident making difficult calls without AI.
What's interesting is that we're beginning to see the same pattern emerge from very different directions.
Harvard Business Review recently warned that organisations risk a gradual knowledge decay when AI-generated work isn't properly challenged and verified.
It's not a dramatic failure.
It's a slow erosion.
The technology evolves quickly.
The behavioural effects emerge much more quietly.
The encouraging part is that behaviours can be changed.
Not through banning AI.
Not through writing longer policies.
But through deliberately redesigning how work happens.
Simple questions at the end of an important piece of AI-assisted work.
What assumptions haven't we challenged?
What still needs human judgement?
If the AI is wrong, where are we most likely to feel the consequences?
Small interventions.
But they're aimed at the thing that matters.
Not the technology.
The behaviour.
I don't think the organisations that thrive over the next few years will simply be the ones with the newest models, the biggest budgets or the highest AI adoption.
I think they'll be the ones that become intentional about the behaviours AI creates inside their teams.
Because AI, on its own, doesn't improve organisational performance.
People do.
The organisations that outperform won't necessarily be those using the most AI.
They'll be the ones that know when AI is helping people think better and when polished output has quietly started doing the thinking for them.
That's the work I'm focused on: helping organisations identify the AI-related behaviours that strengthen performance, spot the ones that quietly erode it and intentionally shape the ones that matter most.
I'm curious.
What AI-related behaviours are you starting to notice inside your own organisation?
For anyone interested in the background to this article, here are the two pieces I found particularly thought-provoking:
• AP News coverage of the Deloitte Australia / Australian Government report: https://apnews.com/article/ab54858680ffc4ae6555b31c8fb987f3
• Harvard Business Review - Don't Let AI Slop Muck Up Your Company's Processes: https://hbr.org/2026/06/dont-let-ai-slop-muck-up-your-companys-processes
