By Melissa Cogger, Partner, Bowmans
The deployment of AI in human resources (HR) is not, in itself, unlawful. The South African legislative framework does not prohibit the use of AI; rather, it requires that AI be deployed in a manner that is transparent, fair, accountable and subject to meaningful human oversight. The challenge for employers is to harness the efficient benefits of AI while ensuring compliance with the Protection of Personal Information Act (POPIA), the Employment Equity Act, the Labour Relations Act, and other applicable legislation.
AI literacy for employers and staff
A threshold requirement for responsible AI deployment is AI literacy. Decision-makers, including HR professionals, line managers and members of disciplinary and selection panels must understand what AI is, how it works, what it can and cannot do, and what risks it presents. Staff who interact with AI tools must be trained on their proper use and limitations. An employer who deploys AI without understanding it cannot meaningfully oversee it or explain its outcomes.
Human agency and oversight
The overarching principle is that AI must supplement, not supplant, human judgement. Every material employment decision must be subject to meaningful human oversight. This means:
- AI outputs (such as candidate rankings, performance scores or retrenchment recommendations) must be treated as inputs to a human decision-making process, not as final decisions;
- the human decision-maker must exercise independent judgement and must be empowered to override the AI recommendation where appropriate or to prohibit the use of AI in certain employment-related decisions;
- the human decision-maker must be able to explain the reasons for the decision to the affected individual; and
- if AI is the sole basis on which a decision is made, affected individuals must be given an opportunity to make representations before a final decision is taken.
Practical steps for employers
Drawing on the legislative framework discussed in this series, the AI Policy Framework, and international best practice, we recommend that employers take the following practical steps:
- Ensure board-level governance and accountability: The Companies Act, 2008, read with King V, requires directors to exercise care, skill and diligence when approving, deploying or overseeing AI systems. Directors cannot discharge their duties solely by delegating to technology specialists or technology providers. Boards should establish clear lines of accountability for AI-related decisions, ensure that AI governance forms part of the organisation’s broader risk management framework and receive regular reports on AI deployment, performance and incidents.
- Align AI governance with King V principles: King V Recommended Practice 109(c) requires adherence to the values of ethics, human centricity, accountability, transparency, explainability, security, privacy, fairness and trustworthiness. It also requires clear accountability for AI decisions, actions, outputs and outcomes, including subjecting automated technologies to human oversight and override mechanisms commensurate with the level of risk to the organisation and its stakeholders. Employers should integrate these principles into their AI governance frameworks.
- Develop and implement an AI policy: The policy should set out the permissible and impermissible uses of AI in the workplace, define roles and responsibilities, establish governance structures, and set out the principles that will guide AI deployment (transparency, fairness, accountability, human oversight).
- Vet AI providers thoroughly: Before procuring AI tools, employers should interrogate the vendor’s privacy policies, data security measures, bias mitigation protocols and the explainability of the tool’s outputs. Prefer paid subscription services that offer contractual commitments over open or free AI tools that may lack adequate safeguards.
- Implement access controls: Not every employee needs access to AI tools. Access should be restricted to authorised users who have received appropriate training.
- Ensure transparency: Employees and applicants must be informed when AI is being used in decisions that affect them (particularly where the decision is solely based on the AI output) and when and why their personal information is being processed by AI. This is both a POPIA requirement and a matter of good practice. Ensure that processing notifications are appropriately updated.
- Conduct impact assessments and ensure there is a personal information compliance framework: Before using AI to process personal information, as a responsible party under POPIA, an employer must ensure that its personal information impact assessment is updated and that its compliance framework is continuously implemented, monitored and maintained in response to AI use. This is to ensure that adequate measures and standards exist to comply with the conditions for the lawful processing of personal information.
- Audit AI tools regularly: AI systems should be regularly audited for bias, accuracy, fairness and compliance. Outcomes should be monitored for patterns that suggest discriminatory impact.
- Establish a cross-functional AI governance committee: Consider establishing a dedicated AI governance committee comprising representatives from HR, legal, compliance, IT, data privacy and risk management. This committee should be responsible for reviewing AI use cases before deployment, monitoring AI outcomes, addressing incidents and complaints, and ensuring ongoing compliance with applicable laws and governance standards; and
- Maintain human oversight at all times: Never substitute AI for human judgement in material employment decisions. AI should enhance the decision-making process, not replace the decision-maker.
The hybrid approach: AI as a complement, not a replacement
The responsible deployment of AI in HR requires a hybrid approach. AI tools are powerful instruments for processing large volumes of information, identifying patterns, and providing data-driven insights. However, they cannot exercise judgement, empathy or contextual understanding in the same way a human can. These are irreducibly human capacities, and they remain essential to fair and lawful employment decision-making.
The key message for employers is one of cautious optimism: AI offers transformative potential for HR efficiency and decision-making quality, but it must be deployed within the bounds of existing legal obligations. Overall, explainability and human oversight are essential. Employers who heed these requirements, consider their legal obligations and who invest in AI literacy, governance frameworks and compliance infrastructure, will be well-positioned to harness the benefits of AI while mitigating its risks.
