Virginia Scully | Human Kind Collective
We have spent years asking what AI can do for our businesses. We are only just starting to ask what it is doing to our people.
The advertising, marketing, and media industries have been early and enthusiastic adopters of AI. Faster briefs. Smarter targeting. Streamlined operations. Leaner teams doing more with less. It feels like it’s all anybody is talking about.
But there is a conversation happening quietly — in HR offices, legal teams, and WHS advisories — that most of the industry has not been part of yet.
I have been working alongside SMEs, building out their people and culture models using AI. Streamlining processes, improving data accuracy, and saving time. It is genuinely exciting work. But the more I do it, the more I keep coming back to the same question nobody seems to be asking.
What happens when the machine gets it wrong? And more importantly, who is responsible when it does?
In December 2025, Victoria's OHS (Psychological Health) Regulations 2025 commenced. They place a clear and enforceable duty on employers to identify, assess and control psychosocial hazards. When AI contributes to a decision that causes psychological harm to an employee, the duty of care does not sit with the tool. It sits with the employer.
In February 2026, NSW went further. The Work Health and Safety Amendment (Digital Work Systems) Act 2026 makes it explicit that employers must ensure their use of AI, algorithms, automation and online platforms does not put workers' health and safety at risk. Unions now have expanded powers to inspect digital work systems where a WHS breach is suspected.
The way your managers use AI is no longer just a productivity question. It is a workplace health and safety one.
Three risks hiding in plain sight
First, intrusive monitoring is now a named psychosocial hazard. Productivity tracking, screen capture, keystroke logging — if your organisation uses these tools, they need to be assessed as safety risks. Research shows the impact is even more pronounced for marginalised cohorts.
Second, opaque decisions create legal exposure. When employees cannot understand why an AI-assisted process produced a certain outcome — and cannot meaningfully challenge it — that is a recognised psychosocial hazard called poor organisational justice. It is not just a culture problem. It has legal weight.
Consider this scenario. A mid-sized firm uses an AI tool to score employee performance across the year. At review time, one of their best people — a woman who has been on parental leave for three months — receives a lower score than her peers. The tool weighted her output during that period the same as everyone else's. No one in HR noticed. She was not told how the score was calculated. She did not get the pay increase she was expecting.
She raises a formal complaint. The firm cannot explain the methodology. They cannot show a human reviewed the output before it influenced her outcome. That is not just a bad day for HR. That is a psychosocial hazard, a potential discrimination claim, and a star employee walking out the door.
Third, bias in your tools is your problem. If an AI tool is trained on historical data that disadvantages certain groups and it influences a people decision, not knowing is not a defence. Foreseeable harm is exactly what these laws are designed to prevent.
The numbers back it up
Australian research across 31 organisations found that psychological harm — from job insecurity, work intensification and unclear role changes — is the most likely consequence of poorly managed AI adoption.
More than one in five small business owners are unaware they are legally responsible for non-physical workplace hazards, including those created by AI tools. And with 92% of CHROs anticipating further AI integration this year, the gap between adoption and accountability is growing.
Three questions worth asking now
Before any AI-assisted people decision, someone in your organisation should be able to answer these: Can a human explain this decision? Can the affected employee understand it? Can they challenge it?
If the answer to any of those is no, the organisation is exposed.
This is the conversation the industry needs to have
The AI adoption plan and the psychosocial risk register should be linked documents, not siloed deliverables sitting in different parts of the business. That connection is the gap most organisations have not closed yet.
Confidence in AI comes from your P & C team and your Managers knowing where the risks are and having a plan.
Virginia Scully is a People and Culture consultant working with SMEs and HR teams on psychosocial safety and responsible AI use in people decisions.