THE KEY INSIGHT: 95% of executives report individual productivity gains from AI. Only 29% are seeing enterprise-wide ROI. The gap between those two numbers is not a technology problem, it is a culture problem. Organisations that treat AI like a traditional systems rollout will fail. The ones winning are those that have rebuilt how they manage change itself.

Every significant technology shift in the past thirty years has followed a predictable pattern. Select the tool. Train the workforce. Stabilise. Move on.

AI does not work that way. And organisations that treat it as though it does are about to learn an expensive lesson.

I have been working with organisations on transformation for twenty years, across culture diagnostics, leadership development, psychosocial risk, and now AI-integrated operational systems. The pattern I keep seeing is the same one Jordan Wilson outlined on the Everyday AI podcast: the organisations in the top five per cent of AI adoption are not winning because they picked better tools. They are winning because they changed how they manage change itself.

Why doesn't the old change management model work for AI?

Traditional change management assumes a destination. Here is where we are. Here is where we are going. Learn the new system, practise, stabilise, done.

AI does not give you a destination. It gives you a moving target.

The tool that delivers massive productivity gains today will still exist in a month, but there will probably be something faster sitting beside it. The workflow you redesigned in Q1 may need redesigning again in Q3, not because it failed, but because the capability set shifted underneath it.

This is what I call the relearn loop. It is not learn, practise, master. It is learn, apply, consider other use cases, remain adaptable, and be ready to relearn. That cycle does not end. And it is fundamentally different from anything we have asked our workforce to do before.

Why isn't AI delivering enterprise-wide ROI?

The statistics from IBM and MIT research tell a clear story. Individual users are finding value in AI almost universally. They are drafting faster, analysing quicker, automating the tasks that used to eat their afternoons. But when you zoom out to the enterprise level, the returns collapse.

The reason is not complicated. It is cultural.

Individual productivity gains happen in isolation. Someone finds a better way to do their work and keeps doing it. Enterprise ROI requires those gains to compound across teams, functions, and workflows. That only happens when people are collaborating, sharing what they have found, testing each other's approaches, and building on each other's solutions.

Most organisations are not set up for that. They are set up for individual performance, competitive recognition, and siloed accountability. Someone learns a new AI capability and it becomes an opportunity to demonstrate personal value rather than a shared resource. Two weeks later, someone else discovers a more effective approach, but there is no structure for that exchange.

The relearn loop requires collaboration at a scale most organisations have never attempted. Not collaboration as a value on a poster. Collaboration as an operational system, with time allocated, processes defined, and outcomes measured.

What cultural infrastructure do organisations need for AI integration?

When I look at organisations that are making AI integration work at the enterprise level, three things are present that are absent in most others.

First, collective stories are being actively rewritten. Every organisation carries embedded narratives. "IT never listens to us." "Leadership does not communicate." "Nothing ever changes around here." These stories become shared beliefs, and shared beliefs become the foundation of culture. If those stories are not surfaced and challenged, they will quietly sabotage every AI initiative you launch. People will not collaborate on new ways of working inside a culture that has taught them participation is pointless.

Second, the pace of implementation planning has changed. The traditional five-year systems roadmap is gone. These organisations are working in 30 to 90 day cycles, mapping pain points from the ground up, prioritising by which changes meet the most organisational objectives simultaneously, and building in structured review points. The implementation plan itself is designed to be rewritten, because it will need to be.

Third, investment ratios have shifted. There is a rule of thumb emerging that organisations should be spending five to ten times what they spend on the tool stack on training their people, building adaptive capability, and bringing in external expertise to rebuild systems and processes. I am not seeing most organisations do this. But the ones that are getting enterprise-level returns have made this shift, because they recognise that the technology is the easy part.

Is AI a psychosocial risk factor in the workplace?

The most common mistake I see is treating AI integration as a technology rollout with a people-management component bolted on. "Here is the new tool. Here is your training. Off you go."

That framing misses the fundamental nature of this change. We are not asking people to learn a new system. We are asking them to exist in a permanent state of learning, to let go of mastery as a goal, and to find motivation in a process that never fully resolves. That is a psychological shift, not a training exercise.

It also creates real psychosocial risk. Role uncertainty. Skill obsolescence anxiety. The erosion of professional identity when tasks that defined someone's expertise are suddenly automated. These are not hypothetical concerns. They are the lived experience of workers across every sector I operate in, from mining to aged care to government.

Australian WHS legislation is catching up. The NSW Digital Work Systems Act 2026 now requires employers to assess psychosocial risks arising from digital and AI-driven work systems. Victoria's Psychological Health Regulations place explicit obligations on employers to identify and control psychological hazards, including those introduced by new technology and changed work design. If your AI strategy does not account for the human impact, it is not just a missed opportunity — it is a compliance gap.

If your AI strategy does not account for this, it is not a strategy. It is a rollout plan with a gap where the people should be.

How should leaders approach AI change management differently?

The organisations that will see enterprise-wide returns from AI over the next eighteen months are not the ones with the biggest technology budgets. They are the ones where leadership has built a culture that can sustain continuous change without burning people out.

That means a culture of collaboration rather than competition. A culture where innovation is shared rather than hoarded. A culture where the relearn loop is supported by structures, not just goodwill.

If you are in a leadership role right now, the question is not which AI tools to invest in. The question is whether your organisational culture can support the kind of change AI actually demands. If the answer is not yet, that is the work that matters most. And no amount of technology spend will do it for you.