Australia leaders warn AI governance lags adoption

Australian business and health leaders are using AI Appreciation Day to warn that governance and security are lagging behind the rapid adoption of artificial intelligence across key sectors.

They argue that contact centres, hospitals and large enterprises are now on the frontline of an AI shift reshaping risk, accountability and customer trust.

In enterprise security, the spread of agent-based AI across corporate networks is heightening concern about exposure to fast-moving cyber threats. Zak Menegazzi, APJ director, Armis from ServiceNow, said organisations are entering an "agentic era" in which software agents can act autonomously inside business systems while hostile actors deploy similar tools at scale.

Menegazzi said this shift has created a structural tension between the way security architectures were designed and the way AI now operates within them. Traditional network segmentation assumed clear boundaries and tightly controlled pathways. Agentic systems are designed to span many of those boundaries and, once connected, can move laterally with little friction.

That turns the idea of a "flat" enterprise network from a theoretical risk into a live issue. Security teams must build guardrails around agents that can trigger actions across infrastructure, data stores and applications in seconds. The goal is to contain both malicious activity and unintended harm from agents pursuing a task without broader context.

On the defensive side, Menegazzi said cyber teams will need AI-driven detection and response platforms that operate at machine speed, rather than manual workflows built for human-scale incidents. That includes automated exposure management, faster remediation and continuous feedback loops so models can adapt as attackers change tactics.

Contact centre operators are confronting a different side of the same debate as they weigh how far to push automation into customer interactions. "Prime Minister Anthony Albanese's address on AI in Australia's interests today put a sharper point on a debate contact centres have been living in for years: where AI should sit, and where it shouldn't. Contact centres are where AI meets real customers every day, at scale, in real time, so that debate has always mattered here first. The lesson from 2026 so far is clear. AI works best when it takes on volume: routine queries, repetitive tasks, and the interactions a machine can handle well. People still own the moments that need judgement, empathy, or a decision only a human can make. We built NeonNow on that split deliberately, because customers can tell the difference and they reward the companies who get it right. True appreciation of AI means being honest about its limits, as well as its capability. Enterprises earning customer trust are the ones drawing that line clearly. With Australia's national conversation on AI standards now catching up to what contact centres already knew, that discipline deserves marking on a day named for AI," said Mike Powrie, founder and chief executive officer, NeonNow.

Healthcare leaders, meanwhile, place governance at the centre of AI decisions rather than the technology itself. Telstra Health executive Dr Monica Trujillo said hospitals and clinics are experimenting widely with AI tools such as digital scribes and decision-support systems, but only a minority have moved beyond pilots.

She said the main constraint is not algorithm performance but clarity around roles, oversight and incident response when AI influences clinical judgement. Questions about who is accountable when an AI-supported decision goes wrong remain unresolved in many organisations. Those questions become more urgent as tools move from administrative functions into diagnosis, triage and care planning.

Trujillo pointed to emerging models of continuous monitoring, similar to long-standing pharmacovigilance processes for medicines, as a likely template. Health systems will need to track AI-related incidents, assess real-world drift from intended use and adjust governance frameworks as evidence accumulates. That includes close attention to data quality and bias, which can translate directly into clinical risk if left unaddressed.

For now, AI Appreciation Day has highlighted a common thread across sectors. Security specialists, contact centre operators and health leaders are each experimenting with AI at scale, while arguing that success will depend on the strength of the human governance and judgement around it, not the sophistication of the models alone.