AI adoption rarely stalls because people can't use the tools. It stalls because no one has told them it's allowed.

New Australian research backs this up: over 40% of employees say using AI at work feels like cheating, and one in three use it without their employer knowing.

The fix isn't more training. It's leadership going first and making permission explicit.

At an executive roundtable we hosted in Sydney this month, one story kept coming up. A team had spent months circling the same unspoken question: were staff actually allowed to use AI? Nobody had said no. Nobody had said yes either. Employment Hero's Paradox at Work report puts a number on that silence. More than 40% of Australian employees say using AI at work feels like cheating, and one in three are using it without telling their employer. The tools were never the problem. The permission was.

Why does AI adoption stall even when everyone has the tools?

It's tempting to assume the holdouts just haven't caught up yet, that they need a course or a policy document before they'll get on board. But the people avoiding AI at most companies aren't confused about how it works. They're avoiding it in front of their manager.

James Keene, Employment Hero's Asia-Pacific managing director, put it plainly in the same report: "When staff feel they have to hide their AI use, it's rarely because they're trying to break the rules. More often, they're simply unsure whether it's genuinely encouraged." That's not a training gap. That's a permission gap, and it's a leadership problem to close.

What actually happened when one team got a Friday to experiment?

One company at the roundtable had already found their way through it, and not with a training course, a policy, or a memo from the CEO. They ran an internal hackathon on a Friday. Everyone got a small AI budget and a low-stakes challenge. The brief was simply: see what you can build.

By the end of the day, the people who had been quietly avoiding AI were the ones demoing what they'd made.

What shifted wasn't skill. It was permission.

Why is the gap between your sales team and everyone else a warning sign?

At several of the businesses in the room, the sales teams were already moving. Training their own agents, building their own workflows, quietly getting ahead. The rest of the business wasn't. And that gap was creating tension nobody had named out loud: one team racing forward, everyone else unsure whether they were even allowed to start.

The companies that get AI right won't be the ones with the best tools. They'll be the ones who brought the whole team along at the same time.
— Teddy Luck-Moreau

How do you actually give a team permission to use AI?

Not with another rollout plan. The businesses making real progress are doing three things at once: leaders using AI visibly and naming it as encouraged, giving staff a low-stakes way to try it without asking first, and treating an AI-assisted attempt that didn't work as a normal part of learning rather than something to explain away.

Dr Anna Kiaos, a researcher in psychiatry at UNSW, made the same point in the Paradox at Work report: "When leaders use AI openly and show their teams, 'this is how we work now,' the guilt and shame fall away, and the gains start to compound." The number worth remembering is the one we opened with: 40% feels like cheating, one in three hide it. That's not a skills problem. It's a silence problem, and only leadership can end it.

"AI is already making people better at their jobs - more productive, doing higher-quality work and developing new skills. The only thing holding the workforce back now isn't the technology, it's uncertainty about whether employees are actually allowed to use it in the ways that work for them. The workforce has already voted for AI with its behaviour, and now it is leadership's job to make it official by talking about it openly, confidently and with support. When leaders use AI openly and show their teams 'this is how we work now,' the guilt and shame fall away and the gains start to compound."

— Dr Anna Kiaos

This was the clearest thing we heard at the roundtable. The blocker is rarely the tools. It's trust, permission, and leadership going first.

If you want to work out where the permission gap sits in your own team, we've put together a short self-assessment. It takes about ten minutes. Details below.

If you were in the room in Sydney and want to keep the conversation going, you can chat to us here

Frequently asked questions

Is this really about AI skills, or something else?

Something else. Most staff who avoid AI at work already know how to use it. The blocker is uncertainty about whether it's actually sanctioned, not ability.

Do we need a formal AI policy before staff can start experimenting?

No. A policy can come later. What staff need first is a clear, visible signal from leadership that using AI is encouraged, not a document sitting on an intranet nobody reads.

What if our sales team is already ahead and everyone else is behind?

That gap is common and worth addressing directly rather than letting it sit. Bring the rest of the business in deliberately, on the same timeline, rather than leaving other teams to catch up on their own.

Is a one-day hackathon actually enough to change behaviour?

It's enough to change the signal. A single low-stakes, well-resourced afternoon can show a team that experimenting is safe and expected. What sustains the shift afterwards is leadership continuing to use AI openly.

Does giving staff permission to use AI mean less oversight?

No. Permission and oversight aren't in tension. Staff who feel safe being open about their AI use are easier to support and correct than staff who are hiding it.

Where do we start if leadership hasn't used AI themselves yet?

Start there. Staff take their cue from what leaders visibly do, not from what's written in a memo. Leadership using AI openly, including making visible mistakes with it, does more than any training course.