Here’s a number that should be on every leadership team’s radar, and mostly isn’t.
Low AI literacy is costing Australia $18.9 billion in foregone economic growth. Not from failing to buy AI — from failing to use the AI we’ve already bought.
That gap, between what AI can do and what it’s actually delivering, is where 2027 will be won or lost. And for any leader carrying an AI agenda into next year — whatever your function — three forces are about to converge. Individually, each is manageable. Together, they’re an inflection point.
Let me walk you through them.
One: the returns still aren’t showing up
Start with the number I keep coming back to, because it still holds true. MIT’s research found 95% of enterprise AI projects fail to deliver a measurable return — and even the ones that technically succeed rarely hit the ROI their business cases promised.
Now, for a moment, borrow the discipline a good CFO would bring to this. If any other category of capital investment came to your board with a 95% miss rate on expected returns, it wouldn’t survive the first meeting. You’d model it, stress-test it, demand a hurdle rate. Yet AI spend keeps getting waved through — because the fear of being left behind is louder than the discipline we’d apply to literally any other investment.
That’s not sustainable into 2027. At some point — and I think next year is the point — boards stop accepting “everyone’s investing in AI” as a business case and start asking the obvious question: where’s the return?
Two: the literacy gap is a measurable, closable dollar figure
Here’s why the returns aren’t showing up — and it’s not the technology.
New research from RMIT Online and Deloitte Access Economics surveyed over 2,000 Australian workers and found something remarkable. 84% already use at least one AI tool at work. But only 7% have reached advanced AI literacy. More than half are still beginners.
Adoption is not the problem. Adoption already happened. Capability is the problem.
And the gap has a price tag, which is what turns this from a training footnote into a leadership priority. Lifting a worker from beginner to intermediate literacy is associated with roughly a $7,000 annual wage uplift; progressing to advanced literacy, nearly $11,000. Multiply that across a workforce and you see how the national figure reaches $18.9 billion in unrealised growth.
Read those two findings together and the picture sharpens. The 95% of AI projects that don’t deliver, and the 93% of workers who haven’t reached advanced literacy, are the same story told from two ends. The technology isn’t underperforming. The humans using it haven’t been equipped to extract the value — and nobody’s been measuring the shortfall.
Three: shadow AI is the risk sitting underneath both
Now the part that should worry any leader most, because it’s the one you can’t see on a dashboard.
The same RMIT research found that nearly half of Australian workers — 49% — are teaching themselves AI through trial and error. No training, no guardrails, no oversight. Building just enough technical skill to be useful while their critical judgement lags behind.
That’s shadow AI: your people feeding company data, client information and commercially sensitive detail into tools nobody sanctioned, learning by guesswork, with no one checking the output before it shapes a decision. It is, quietly, one of the greatest risks facing businesses today — and it’s a direct consequence of the first two forces. When the organisation doesn’t provide structured capability-building, people build it themselves, invisibly, at exactly the level of judgement you’d least want near your most important work.
This is where the literacy gap stops being an opportunity cost and becomes an actual liability — error, exposure, and decisions made on unverified AI output that no one flagged. It doesn’t matter which function you lead; if your people touch information that matters, shadow AI is already in your building.
What this means for 2027
Put the three together, and the inflection point is clear. The spend is under scrutiny. The capability isn’t there to justify it. And the gap is being filled by unsanctioned, unmanaged use.
The leaders who navigate 2027 well won’t be the ones who spent the most on AI, or the least. They’ll be the ones who did three things deliberately:
Measured the literacy gap in their own workforce, in dollars, the way they’d measure any other capability shortfall.
Closed it with structured, judgement-focused capability-building — not another tool rollout, and not self-guided trial and error, but real practice on real work with real oversight.
Brought shadow AI into the light, by giving people a sanctioned, supported way to build skill so they stop building it in the dark.
None of that is a technology project. All of it is a leadership and capability project — which is precisely why it lands on your desk, whatever your title, and not the CIO’s alone.
The $18.9 billion isn’t lost because Australia bought the wrong tools. It’s sitting unclaimed because we’ve treated AI as something you purchase rather than something your people have to become good at.
2027 is the year that distinction stops being philosophical and starts showing up in the numbers. The remainder of 2026 is when you start to plan.
Which of these three is furthest along in your organization—the spend scrutiny, the literacy gap, or shadow AI? I’d genuinely like to know where leaders think the pressure lands first.
If you’re a leader staring at an AI investment that hasn’t paid for itself yet, this is the conversation I have with executive teams: how to measure your literacy gap in dollars, close it with real capability rather than another tool rollout, and bring shadow AI into the light before 2027 does it for you. That’s the work — and if it’s the conversation your leadership team needs to have, I’d welcome it. DM me
And if you’re planning a 2026–27 conference or leadership event, this is the keynote I’m most asked for right now — the human side of AI adoption, with a live, bespoke AI demonstration that shifts a room from fear to confidence and curiosity to capability .
