Athalie Williams on why AI underperformance is a leadership problem, not a technology one
Australian CEOs, according to recent AFR reporting, are more confident about growth, more eager to invest in AI, and more anxious about falling behind than their global peers. They are also forecasting deeper job cuts. Held together, that combination points to something more complicated than a story about technology replacing people.
It is a story about what happens when leadership behavior outpaces the conditions needed to support it.
When pushing harder is the wrong move
Early in her career, Athalie Williams watched a major technology program drift further off course with each passing month. Deployments were not delivering the promised customer experience. Expected cost reductions had not materialized. The instinct in the room was to push harder — accelerate delivery, invest more, take further cost out of the back office to compensate. Everyone could feel the strain. No one wanted to be the person who said the program was no longer solving the right problem.
It is a dynamic Williams has seen repeat across organizations. A senior leader, after a meeting, puts it plainly: “We all know this isn’t working, but no one wants to be the one to stop it.” The fear of pausing feels greater than the risk of continuing.
Job cuts follow a similar logic. When AI investment is not delivered, reducing headcount appears to do two things at once: it signals decisive action and creates short-term cost relief. What it does not do is address the conditions causing the underperformance. Often, it makes them worse — removing the judgment, institutional memory, and contextual knowledge that better execution depends on.
Three gaps that headcount reduction cannot close
The AFR reporting notes that only a small minority of Australian CEOs have seen financial returns from AI so far, yet many are planning deeper cuts while expressing concern they are not moving fast enough. That combination of speed and uncertainty is where the real risk sits.
Across her career, Williams has found that AI underperformance has rarely been a capability problem first. More often it comes back to clarity, coherence, or foundations — frequently all three.
Clarity means defining the problem with precision before committing to a solution. It means articulating what value is at stake, protecting the investment once committed, and being willing to stop work that is not delivering — even when sunk cost or organizational pride makes that uncomfortable. When clarity is treated as an aspiration rather than a working discipline, organizations move quickly but without the alignment needed to create value.
Coherence is harder to see. It is the alignment of purpose, priorities, behaviors, and how decisions actually get made. When it is present, direction is clear, activities move the needle, and leaders reinforce the same priorities rather than inadvertently working against each other. When it is missing, the effects show up in duplicated effort, inconsistent signals, and teams that are busy without building toward anything. Enthusiasm gets mistakes for readiness. Early experimentation gets treated as proof of scalability.
Foundations are the unglamorous piece. Well-designed initiatives still struggle when built on old systems, scattered data, and processes never intended to work with AI. Meaningful progress typically requires sustained investment over 18 to 24 months — and that timeframe reflects organizational work, not just technical implementation. When leaders underestimate this, they overpromise. When they overpromise, cutting headcount becomes the most available lever.
What boards should be asked
AI has become a governance topic, and boards have a responsibility here that goes beyond approving investment cases.
The useful questions are not about whether the organization is moving fast enough. They are about whether it is moving with sufficient clarity. What problem is being solved? What is the value logic underpinning the investment? Does the organization have the maturity to execute at the scale being planned? Are job cuts addressing a genuine productivity gain, or filling a gap left by overpromising?
There is also a question board tending to be underweight. Judgment, context, and institutional memory are not soft concepts. They are part of the fabric that makes coherent execution possible. When leaders treat them as an HR matter to address later, AI initiatives drift. The organizations making genuine progress build human capability alongside the technology, not as an afterthought.
Before the lever gets pulled
There is a version of the current moment that ends well for Australian organisations. CEOs are genuinely invested. Boards are increasingly engaged. Ambition is not the problem.
The risk is that pressure to show results — combined with the difficulty of early returns — produces a pattern of cutting capacity at the very point when judgment, coherence, and clear thinking matter most. Boards, executive teams, and management have the ability to make a different choice. The question is whether they will recognize the moment before the lever gets pulled.
