Beyond an AI policy: helping employees know when to trust AI

As workplace trust in AI grows, HR leaders have a critical role to play in helping employees question outputs, apply judgement and know when humans need to step in.

Trust in AI is growing among Australian employees. On the surface, that’s encouraging. It suggests efforts to encourage experimentation are working and employees are seeing genuine value in the tools they’re using.

ELMO’s latest Employee Sentiment Index (ESI) found 72% of Australian employees now use AI at work. Among those who would turn to AI for workplace advice, 44% say better-quality answers are one reason they would choose it over a person.

But the research suggests confidence in AI may be developing faster than the rules around it. The same research found 43% of employees don’t think it is clear when AI should and shouldn’t be used in their workplace.

For HR leaders, that raises a different challenge. As AI becomes a normal part of work, employees need to know when to trust its output, when to question it and when a human needs to make the final call.

For ELMO Software, trustworthy AI means keeping a human firmly in the loop.

“AI can analyse enormous amounts of information and surface insights far faster than a person could, but that doesn’t mean we should hand over human judgement. In my team, we see this play out daily – AI models flag patterns and generate recommendations, but a person still has to interrogate what they're seeing, apply context, and own the decision that follows," says Ramesh Thiagalingam, VP of MLOps and Platform Engineering at ELMO Software.

"That principle is baked into how we build AI at ELMO. It's part of why we've pursued certifications like ISO 42001 for responsible AI management. AI should make us faster and sharper, not replace the accountability that comes with being human."

AI literacy requires a healthy dose of scepticism

Interestingly, the more sophisticated an AI user is, the more – rather than less – likely they are to have felt uncomfortable with how AI is used at work.

The ESI found 59% of regular AI users have experienced discomfort with its use, compared with 36% of non-users.

That discomfort isn’t necessarily a bad sign. More experienced users may simply have a better understanding of where AI performs well and where its limitations begin.

“Being confident with AI shouldn’t mean accepting every output at face value,” Thiagalingam says. “Models can get things wrong. They can introduce bias or produce an answer that appears convincing without being accurate. Good AI capability includes knowing when something needs to be challenged, validated or handed back to a person.”

This is becoming more important as AI shifts from experimentation to an expected workplace capability. Almost two-thirds (65%) of Australian employees now consider AI an expected skill at work, while 58% believe employees who don’t use it risk falling behind.

For HR, the challenge is therefore twofold: helping employees participate in that change while giving them the skills and guardrails to do it responsibly.

Capability can’t be a one-off exercise

ELMO Software Chief People Officer Anne Tosky says the speed of change means traditional approaches to workplace training are also having to evolve.

“You can’t treat AI capability as a training program you roll out once and consider finished,” Tosky says. “The technology is moving too quickly, and the way people are applying it to their work is changing just as fast.”

ELMO’s own approach has evolved as employees have become more confident with AI. Early efforts focused on creating opportunities for people to experiment and understand what the technology could do. The focus is now increasingly on practical implementation and helping teams identify where AI can improve the way work gets done.

“We’re moving beyond simply encouraging people to use AI and looking much more closely at workflows,” Tosky says. “How can an individual team build an agent or redesign a process that adds value for everyone? And what capability do people need to use those tools confidently and responsibly?

“That requires ongoing dialogue. People need to be able to raise concerns, share what isn’t working and ask questions without feeling like they’re somehow falling behind.”

That point matters when almost half (49%) of Australian employees say they have felt uncomfortable with how AI is being used in their workplace.

Trust needs to be earned

None of this means organisations should retreat from AI adoption. The ESI suggests employees already recognise its potential: two-thirds believe AI is helpful to their role, with speed, convenience and quality the top reasons employees are increasingly turning to it for workplace advice.

The next phase is ensuring growing confidence is matched by judgement. That means moving beyond simply providing access to AI tools or writing an AI policy. Employees need approved environments in which to experiment, clear expectations about appropriate use, and the capability to interrogate outputs rather than defer to them.

As AI becomes embedded in more workplace decisions and workflows, the measure of an AI-ready workforce won’t simply be how many employees use it. It will be whether they know when to trust it – and when not to.

Download ELMO Software’s latest Employee Sentiment Index to explore how Australian employees are responding to AI and the other forces reshaping work.

Want to see more stories from trusted news sources?
Make HR Leader a preferred news source on Google.