Here’s the deal: AI adoption needs a rebrand.

Everyone’s treating it like another technology rollout where you choose a tool, grant access, publish a policy document, book a training session, and the job’s done.

Except it really isn’t.

Our latest research shows that AI has crossed from experimentation into operations. Fifty-three percent of organisations now have AI implemented in workflows or embedded in strategy, up from just 13% a year ago, and informal experimentation has fallen from 48% to 15%.

That’s a serious shift in 12 months. But having AI in the business isn’t the same as having people adopt it well.

The more AI becomes part of everyday work, the less its success depends on access to the technology and the more it depends on behaviour: whether people know where to use it, when to question it, what information is safe to share and how to turn a useful experiment into a repeatable way of working.

AI adoption is now a people strategy challenge.

Access is easy. Adoption is the harder bit.

Almost half of the 1,505 business leaders and technology professionals we surveyed use AI daily, while 71% use it at least a few times a week. Only around 6% never use it.

On the surface, that looks like adoption.

Look a little closer and the picture becomes less convincing. Only 36% describe themselves as guiding AI to produce stronger outputs. Just 6% are using it in repeatable workflows, and 8% have connected it to tools or systems.

In other words, widespread use doesn’t mean widespread advanced capability.

People can use the same AI tool and get wildly different results. One person uses it for a quick first draft. Another builds it into a workflow, validates the output and improves how the work gets done. The difference isn’t access but confidence, context, judgement and practice.

This reflects a much broader transformation challenge. The World Economic Forum found that skills gaps are the leading barrier to business transformation, cited by 63% of employers globally.

The technology may be moving fast, but organisational capability still has to be built one team, role and workflow at a time.

A clear policy doesn’t guarantee clear behaviour

Most organisations appear to have made progress on the governance basics. Seventy-nine percent of respondents say their organisation’s AI rules are at least somewhat clear, including 41% who describe them as very clear.

But only 16% are very confident that people know what data can and cannot be entered into AI tools. Nearly one in four say their organisation has never provided AI training or a policy refresh.

That gap raises some eyebrows because the risks are already showing up in everyday behaviour:

• 57% identify relying on AI outputs without checking them as a leading risk.
• 57% flag entering confidential or client data.
• 47% are concerned about decisions being made without human oversight.

While a policy can define the rules, it can’t solely coach someone through a great-area decision in the middle of a busy working day.

As Sarah Blanchard, Head of Talent Advisory at Solve, explains:

“Policies create clarity, but they don’t create consistent behaviour. Sustainable AI adoption comes from embedding good practice into everyday work through clear use cases, workflow guidance and regular reinforcement, so the responsible choice becomes the easiest choice. An AI framework should be the operating system for adoption, not a governance document that sits on the intranet.”

The practical question isn’t whether the policy exists, it’s whether people can apply it when the answer isn’t obvious.

Managers are carrying the adoption gap

Managers sit exactly where AI ambition meets day-to-day reality.

They’re expected to interpret organisational objectives, manage quality and risk, answer questions, coach new behaviours and decide when human judgement needs to take over. And most often, they’re doing all of this while their own role is changing too.

While 76% of managers are considered at least somewhat prepared to manage AI use, only 25% cite they’re very prepared. Their hardest challenges include data security and compliance, managing poor or excessive use, knowing when judgement is required and measuring productivity gains.

Curiously, training the team ranks last.

And that doesn’t necessarily mean managers think training is easy. It might instead mean that managers expect someone else to own it.

The rise of dedicated AI enablement roles supports that theory. Microsoft’s Work Trend Index found that 35% of managers were considering hiring AI trainers to guide adoption. But the same research found that 51% believe AI training or upskilling will become a key responsibility for their own teams within five years.

Specialist support can help establish the framework, tools and curriculum. What it can’t do is replace the manager’s role in reinforcing how AI should be used in the actual work.

As Cameron Robinson, Head of Enterprise at Solve, puts it:

“It’s interesting that ‘training the team’ ranks lowest, because stronger training would likely help ease many of the challenges above it—from data security and compliance to overuse, poor use and knowing when judgement is required. We’re seeing more AI enablement roles emerge but relying on one person or team to carry that responsibility is risky; managers are still the critical link between company objectives and day-to-day execution.”

An AI Enablement Lead simply can’t sit beside every employee, review every output or turn every lesson into a new team habit.

So, if managers are the expected distribution network, they need to be equipped accordingly.

That support also can’t stop once the policy is published or the initial training is complete. As Emily Zhang, Director of The HumAIn Impact, observed during our recent After the All-Hands panel on building trust, adoption and lasting change, launch is only a milestone. For AI, the real work happens afterwards, when people encounter unclear use cases, questionable outputs and decisions that require human judgement. Role-based practice, regular feedback and manager-led reinforcement are what turn initial uptake into lasting behaviour.

Building an operating system for AI adoption

An effective AI framework should connect workforce strategy to employee behaviour.

It should tell leaders where AI supports business priorities, give teams approved use cases, establish clear data boundaries and define where human oversight is non-negotiable. It should also address capability, role design, manager expectations and how value will be measured.

Without that link, organisations risk operating with two versions of AI adoption:

The official version, where governance is clear and the technology is delivering value.

And the everyday version, where employees are experimenting independently, managers are making judgement calls without enough support and nobody is entirely sure whether productivity has genuinely improved.

AI adoption doesn’t get stuck because people are unwilling to use the technology. The data suggests they’re already using it at speed.

The challenge is turning that activity into safe, repeatable and valuable ways of working, and giving managers and employees the practical support to make those behaviours stick.

Our research explores how more than 1,500 professionals across Australia and New Zealand are navigating that gap, from policy and training to manager readiness, trust and workforce design. It provides a clearer view of what organisations need to address as AI moves from individual experimentation into everyday operations.