Australian businesses are adopting AI far faster than they are learning to manage it. A new ISACA poll found that 90 per cent of organisations believe their employees are already using AI tools, yet only 22 per cent say the returns have met or exceeded expectations, according to ARN. The gap between use and value has become a governance problem, not just a technology one.
More striking still, 56 per cent of respondents could not say how long it would take to shut down an AI system during a security incident. For boards that remain legally accountable for what their systems do, that is a difficult number to sit with.
What the ISACA poll found
The 2026 AI Pulse Poll asked risk and security professionals how AI is actually being used inside their organisations. The headline figures point to enthusiasm running ahead of control: near-universal adoption, modest measurable returns, and widespread uncertainty about basic incident response.
Jamie Norton, ISACA vice chair and chief information security officer at the Australian Securities and Investments Commission (ASIC), said smaller and mid-sized organisations were struggling the most. Many lack the in-house expertise to assess AI risk, let alone govern it.
Why the gap keeps widening
As internal readiness lags, responsibility is shifting outward. Norton noted that customers increasingly expect their resellers and managed service providers to supply the governance and risk controls they cannot build themselves. That only works if those providers bring a genuine AI governance lens, rather than reselling tools and hoping for the best.
The fundamentals have not changed either. Patching, multi-factor authentication and clear data handling still do most of the heavy lifting. Bolting AI onto weak foundations simply widens the exposure, which is where a structured AI strategy matters more than any single tool choice.
Why Australian businesses should pay attention
The local context sharpens the stakes. The Privacy Commissioner and the OAIC are expected to keep clarifying how privacy and data-protection duties apply to AI, including where data is sent and how inputs are handled. Boards and executives retain accountability for breaches regardless of which vendor was involved.
For organisations handling personal or regulated data, mapping AI use against the Privacy Act compliance requirements is no longer optional, and it is where many Australian operators have the widest gaps.

Closing the AI governance gap
The poll is less a warning about AI itself than about adopting it without a plan. The organisations seeing returns tend to be the ones that decided up front which tasks AI should touch, who owns the output, and how it gets switched off if something goes wrong.
A short AI readiness review is usually enough to surface the obvious gaps before they turn into incidents. For Australian businesses, the next phase of AI is not about adopting more of it, but about governing what they already have.
