International executive and former CHRO Athalie Williams on why the wrong framework is being applied to one of the most consequential questions in Australian business
Artificial intelligence is now a fixture on Australian boardroom agendas. The question is whether it belongs where most boards have put it.
Across the country, directors are approaching AI governance through existing frameworks: cyber risk, data privacy, IT oversight. It is an understandable instinct. These frameworks exist, they are familiar, and AI carries genuine technology risk. But Athalie Williams, a transformation executive who has spent more than three decades leading complex organisational change across resources, telecommunications, and financial services, argues that filing AI under technology risk is precisely the wrong container.
“Boards are asking the right question about how to govern AI responsibly, but they’re often asking it in the wrong room,” she says. “The real questions here are human performance questions, culture questions, and governance questions. Those require a different kind of board attention.”
The framework mismatch
Technology risk frameworks are designed to manage known failure modes: breaches, outages, compliance gaps. They are built around containment. That lens captures only part of what AI governance actually requires.
The more consequential questions are not about what happens when AI fails, but about what happens when it works. How it reshapes decisions, who remains accountable for them, and whether the human judgement that underpins long-term organisational performance is being preserved or eroded.
Williams has described what she calls the “productivity paradox” of AI adoption: organisations invest heavily in the technology, throughput rises, but resilience, judgement, and trust can fall behind if boards and executives do not design for them deliberately. “Faster processes and smarter models don’t automatically make wiser organisations,” she says. “The gap is in the human architecture around the technology, and that gap is a governance question.”
What boards are missing
The tendency to treat AI as a technology investment with associated risk is leading boards to overlook three areas that will ultimately determine whether AI delivers sustainable value.
The first is human capital. Most AI governance conversations at board level focus on what the technology can do. Fewer focus rigorously on what the workforce needs to do differently as a result. Not just in terms of skills training, but in terms of how roles are redesigned, how accountability is structured, and how leaders are equipped to oversee AI-assisted decisions rather than simply execute them.
Williams draws on her experience leading large-scale capability transformations to make the point. Building genuine AI readiness across an organisation looks far less like a training programme and far more like a sustained shift in how the organisation thinks about human capability and its relationship to automated processes. “Boards should be asking their executives not just what AI we’re deploying, but how we’re designing the human layer around it,” she says. “That question is largely missing from governance conversations I observe.”
The second is decision architecture. As AI takes on more of the analytical and procedural work that humans have historically performed, boards need to be explicit about which decisions must remain human-led, which can be hybridised, and which can be delegated to automated systems. Board-level deliberation is required here, not delegation to the CIO or a risk sub-committee.
“High ambiguity, high consequence decisions should be human-led by design,” Williams says. “But most organisations haven’t done the work of mapping their decision landscape in those terms. Boards are approving AI investments without a clear view of what they’re implicitly delegating.”
The third is culture. AI adoption does not happen in a cultural vacuum. How an organisation’s people respond to automation, whether they engage with it, work around it, or quietly resist it, is shaped by leadership behaviour, psychological safety, and the signals boards and executives send about what is valued. Boards that treat AI as a technology procurement question will miss this dimension entirely.
A more useful governance posture
None of this requires boards to become AI experts. It requires them to ask different questions and hold their executives accountable for different answers.
Williams suggests a practical starting point: identify one material decision in the organisation where AI played a defining role in the past quarter and bring it to the board as a learning case. Where was human judgement applied? Was it effective? How did the decision land with employees or customers? Did it reflect the organisation’s values in practice?
“That kind of deliberate review builds board fluency without requiring technical expertise,” she says. “It also sends a signal to the executive team about what the board considers important. That signal matters.”
The deeper shift is treating AI governance as a human performance question rather than a risk management exercise. That means workforce planning, capability investment, and culture stewardship sitting alongside technology oversight in how boards think about their responsibilities.
Australian boards have a reputation for pragmatism. The AI governance challenge is a practical one: the frameworks being applied were not built for the questions now being asked. For organisations where AI is already reshaping how work gets done, updating them is fast becoming an operational necessity.
Follow our WhatsApp channel
