The AI Productivity Paradox
KPMG Australia dropped a report this week that should make every business leader in the region sit up. The headline numbers:
- 74% of Australian organisations are actively deploying AI for productivity gains
- But 87% lack a responsible AI governance framework
- 63% have no employee AI literacy or training program
- 52% can’t identify which of their systems use AI
In other words, nearly three-quarters of Aussie businesses are using a tool they don’t understand, without guardrails, and without training their people — and they’re calling it a productivity strategy.
This is the AI equivalent of handing everyone a chainsaw and telling them to “just be careful.”
📋 The KPMG Framework: What ‘Responsible AI’ Actually Means
KPMG’s report breaks responsible AI governance into five pillars:
- Accountability — Who owns AI decisions and outcomes in your organisation?
- Transparency — Can you explain how your AI systems reach their conclusions?
- Fairness — Are you testing for bias, particularly against protected groups?
- Reliability — Do you have monitoring in place for when AI systems behave unexpectedly?
- Privacy — Are you protecting the data that flows through third-party AI tools?
The vast majority of Australian businesses, KPMG found, are scoring poorly on all five.
💡 The Productivity Trap
Here’s the tricky part: these same businesses are seeing real productivity gains.
Roughly 68% of early AI adopters in the report reported measurable efficiency improvements. Customer service AI agents, code generation tools, document automation — they’re delivering. The numbers are real.
But KPMG’s warning is that these short-term gains are hollow if they’re built on an ungoverned foundation. The report draws a direct line from today’s governance gaps to tomorrow’s:
- Regulatory fines (EU AI Act extra-territorial reach, Australian privacy reforms)
- Reputational damage (AI bias scandals, customer data misuse)
- Legal liability (who’s responsible when an ungoverned AI makes a bad decision?)
- Operational risk (AI drift, hallucinations, model collapse)
“Organisations that treat AI governance as an afterthought will find their productivity gains eaten by regulatory costs within 18 months,” the report concludes.
🇦🇺 Australia’s Regulatory Context
This report drops at a particularly charged moment for Australian AI regulation:
- Privacy Act review — Major reforms in progress, with AI-specific provisions
- Mandatory AI guardrails — The government is consulting on mandatory safety requirements for high-risk AI
- EU AI Act extraterritoriality — Australian companies serving EU customers are already in scope
- NZ alignment — Trans-Tasman regulatory harmonisation is under discussion
The window for voluntary governance is closing. Mandatory frameworks are coming, and companies that haven’t built the muscle will face a scramble.
🌏 NZ Parallels
KPMG’s findings apply equally to New Zealand — arguably more so.
- New Zealand’s AI adoption rates are similar to Australia’s (73% of NZ businesses using some form of AI in a March 2026 MBIE survey)
- NZ’s regulatory framework is less developed than Australia’s — the NZ AI Bill is still in consultation phase
- The talent pool for AI governance specialists in NZ is tiny. Even if a business wants to do the right thing, finding someone qualified to build the framework is genuinely hard
- But the risks are identical: EU AI Act applies to NZ companies serving European customers, Australian privacy reforms have implications for trans-Tasman data flows
For Kiwi businesses, KPMG’s report should be read as a three-year warning: get your AI governance sorted now, while there’s still time to do it at your own pace.
🤔 My Take: Governance Isn’t Glamorous, But It Pays
I’m going to say something that might be unpopular in the “move fast with AI” crowd: basic governance is the highest-ROI AI investment you can make right now.
Not the flashiest AI model. Not the most ambitious deployment. Just knowing what AI you’re using, who’s accountable for it, and whether it could blow up in your face.
Every AI horror story of 2025–26 — the biased hiring models, the hallucinated legal citations, the customer data sprayed across third-party APIs — traces back to a governance gap. Someone deployed AI without asking the basic questions.
KPMG is right to flag this. The firms that build governance muscle early won’t just avoid disaster. They’ll be the ones regulators trust, customers choose, and insurers underwrite.
The rest? They’ll be the case studies in next year’s report about what not to do.
🔍 THE BOTTOM LINE: 87% of Australian businesses lack basic AI governance while rushing to deploy productivity tools. The NZ picture isn’t better. KPMG’s warning is clear: fix the foundations now, or watch regulatory costs eat your productivity gains. Boring governance work is the most important AI investment you can make.
