AI and Human Resources: Managing Opportunities and Risks

Artificial intelligence is fast becoming an everyday business tool, offering employers genuine efficiency gains across workforce functions. However, the pace of adoption means legal and compliance risks, particularly in the context of human resources functions, can easily be overlooked.

Employers should establish a clear AI policy before deploying any AI tools. The policy should identify approved tools, specify what information may and may not be entered (in particular, employees should not input personal information or confidential information into publicly available AI tools), require appropriate human oversight and responsibility, set out consequences of non-compliance, and require that employees check and take responsibility for AI outputs.

Privacy must be considered at this early stage, including whether data entered is ‘sandboxed’ or may be used to train the AI. Privacy policies should also be updated to reflect the organisation’s use of AI.

Across all Australian jurisdictions, work health and safety legislation creates an obligation to ensure the health and safety of workers, and expressly includes psychological health. A poorly managed AI rollout or framework for its use can create psychosocial hazards, for example, by radically changing job content, reducing workers’ autonomy, intensifying workloads, or creating uncertainty about job security. A failure to consult workers about AI implementation is itself a potential psychosocial hazard.

Persons conducting a business or undertaking must eliminate or minimise psychosocial risks so far as is reasonably practicable. Businesses should implement reasonable control measures to mitigate this risk, assess the implementation of AI through a psychosocial lens, communicate clearly with workers about what will and will not change, and continue to monitor impacts after the technology is in place.

AI can assist with drafting documents such as job descriptions, policies, and procedures. However, although AI-generated content may look polished, it can often contain legal inaccuracies, irrelevant material, or information from the wrong jurisdiction. AI can be a helpful tool for these functions, but it is vital to critically review its outputs and ensure it is approved by someone who understands the relevant legal framework.

AI recruitment tools can screen applications at scale, but businesses must beware of algorithmic biases in relation to all kinds of discrimination, but especially age, disability, sex and race. Relevantly, adverse action protections under the Fair Work Act 2009 (Cth) can extend to prospective employees in relation to workplace rights and protected attributes. Employers should test AI tools for bias, maintain meaningful human involvement in hiring decisions, and consider whether the extent of AI use means it should be disclosed to candidates.

Upcoming amendments to the Privacy Act 1988 (Cth), effective 10 December 2026, will require organisations covered by that Act to address automated decision-making in their privacy policies where automated decision-making processes use personal information and could significantly affect individuals’ rights or interests.

There are many potential AI applications in relation to workplace surveillance to increase worker safety, reduce managerial oversight required and monitor outputs. However, it is vital to ensure that appropriate consultation is carried out (see further below) and that privacy considerations are properly addressed.

Bunnings’ recent appeal against the Privacy Commissioner’s determination in relation to facial recognition technology is a timely reminder to ensure that surveillance technology is both necessary and proportionate. The appeal upheld key parts of the Commissioner’s decision in the first instance, including that Bunnings had failed to have adequate signage disclosing the use of facial recognition technology.

Employers considering AI-powered surveillance should conduct a privacy impact assessment to ensure the usage is appropriate, and that all privacy related obligations will be properly satisfied in relation to consent and collection. State-based legislation such as the Surveillance Devices Act 2016 (SA) may impose additional requirements in relation to audio and video surveillance in certain contexts.

AI rostering tools carry risks of inadvertent discrimination. For example, it may allocate fewer shifts to workers with caring responsibilities. These tools may also fail to account for penalty rates, minimum engagement periods, or break requirements under the applicable modern award. If an employer cannot explain how AI allocated shifts, defending a workplace dispute becomes very difficult. Employers should audit rostering outputs regularly, maintain override capability, and ensure the system reflects current award and enterprise agreement requirements. Changes proposed to regular rosters or ordinary hours also trigger a specific consultation obligation, which consultation may be missed if there is no oversight to notice that a change to regular arrangements has occurred.

The Fair Work Act 2009 (Cth) requires that employees covered by a modern award or enterprise agreement must be consulted with about major workplace changes likely to have a significant effect on them, which may include changes to required skills, or the elimination or redesign of jobs. Consultation also must occur where there are changes to regular rostered hours or shifts, which changes may arise from the use of AI rostering tools.

Employees must have a genuine opportunity to be heard and have the potential to influence the decision. Failure to consult may be a breach of the relevant award or enterprise agreement and can defeat a ‘genuine redundancy’ defence in unfair dismissal proceedings.

Employers remain accountable for AI-assisted decisions and must maintain meaningful human oversight. Obligations in relation to consultation, privacy, and workplace health and safety can arise from the point that a decision is made to introduce AI. Employers must strike a balance between leveraging the opportunities presented by AI whilst not falling foul of the many risks that can arise from poor implementation. Ultimately, whilst AI may help to streamline and carry out some human resource functions more efficiently, it cannot and should not replace human oversight.