Millions of Australian workers now secretly managed by AI, new report finds

Workplace software unicorn Deputy has unleashed a new AI agent that can roster staff and manage timesheets in minutes, but the firm has warned about getting carried away with the tech and losing human touch.

Deputy chief technology officer Ciaran Hale said that the biggest obstacle for enterprise adoption is often not technological capability, but customer trust.

Nearly half of Australia’s frontline workers are unaware that artificial intelligence is now controlling core aspects of their jobs, revealing a crisis of trust as technology transforms the way cafes, supermarkets and hospitals manage staff.

This “clandestine deployment” of AI – where efficiency has been prioritised over employee awareness – is the two-front challenge facing Australian software unicorn Deputy, which recently launched a powerful autonomous platform designed to serve its 2.7 million shift worker users.

The firm, which holds a market value of more than $1bn, has recently unveiled its Deputy AI platform in partnership with Amazon Web Services (AWS).

Built on the AWS Bedrock technology stack, the “agentic AI” is designed to automate labour-intensive functions like rostering and timesheet approval, promising to slash the time it takes a small cafe manager to complete a weekly roster from up to 30 minutes to less than five.

The system can process complex, conversational commands, such as: “Sorry, Jared shouldn’t be on this day, can you move him to next Thursday?”

It comes as Deputy also released a study that exposed a deep-seated disconnect between corporate AI adoption and employee awareness. The report revealed a startling “awareness gap”, with 49 per cent of Australian frontline workers unsure or completely unaware that AI is now controlling core aspects of their jobs. Only 16 per cent of employees reported their employer is completely transparent about AI usage, suggesting that for millions of workers, the shift has been a “clandestine deployment” prioritising efficiency over human trust.

“AI isn’t a plug-and-play solution – it’s a culture shift,” Deputy chief executive Silvija Martincevic said. “Our research shows nearly half of workplaces are already using AI, but only one in four workers personally interact with it, and most aren’t told how it affects their jobs. That’s not innovation, that’s isolation. The real challenge isn’t building smarter tools, it’s building trust.”

Deputy chief technology officer Ciaran Hale said that the biggest obstacle for enterprise adoption is often not technological capability, but customer trust, citing public anxiety around automated processes.

To counter the fear that an autonomous system might “go rogue” or be misused, the company has built its platform on a stringent, multi-layered framework of guardrails inherited from the AWS security ecosystem.

The first line of defence ensures the AI agent inherits the human user’s permissions, meaning it cannot perform an action, such as rostering an employee, if the logged-in manager does not have the administrative privilege to do so manually.

The second, more novel layer of security, is AWS’s Amazon Bedrock Guardrails. This system operates as a firewall for the large language models (LLMs) that power the agent, controlling the agent’s scope and preventing misuse. Mr Hale said that Guardrails stops “people doing things like just using the agent as their own personal chat GPT”. The system is configured to flag and reject off-topic or overly broad queries, ensuring the agent remains strictly focused on workforce tasks.

“Amazon Bedrock Guardrails allow us to control misuse of our underlying APIs and reduce hallucinations, helping us deliver a safer product,” Mr Hale said.

While the efficiency metrics of the platform are clear, the workforce anxiety is focused not on job displacement – 59 per cent of workers were not worried about being replaced – but on being left behind.

A vast majority, 81 per cent, stated they would be more willing to embrace AI if they were simply given proper training and preparation.

Ironically, the drive toward automation is creating a demand for more, not less, human intervention at the corporate level. Mr Hale said that the company is now focused on “upskilling our support team in AI”, recognising that a human with native AI knowledge is often required to diagnose an error and rebuild trust when an agent misinterprets a complex prompt.

The company’s own report found that two-thirds of shift worker roles require a human or personal touch, with 94 per cent agreeing that compassion, emotional support, and empathy will always remain uniquely human strengths.

The broader rollout of the Deputy AI platform is anticipated to begin in March or April next year.