Waar gaat deze meeting over?
Beyond Responsible AI principles toward a workable HR AI governance mode
As AI moves into real HR decisions, governance needs to become practical. In this sparring session, we’ll explore what a workable HR AI governance model looks like in practice, using cases such as job grading, payroll, recruitment and performance.
We’ll discuss who decides, what gets documented and reviewed, how AI is tested, and where the red lines are. Sandrine Bardot will bring her perspective to the discussion, helping us explore how HR can move from AI experimentation to governed, responsible AI-enabled decision-making
Presentatie:
Key take-aways:
The session with Sandrine Bardot, representing The Bardot Group, explored how organizations can move from Responsible AI principles to practical governance when AI begins influencing consequential HR and Rewards decisions. The central challenge was straightforward but significant: who actually controls the decisions AI is influencing?
The discussion highlighted that simply having a human in the loop does not, by itself, resolve questions of accountability. As AI moves from Inform → Challenge → Recommend → Decide, organizations need to consider decision rights, accountability, guardrails, explanation, and visibility into what is happening “under the hood.” The meeting also identified the EU AI Act and governance of AI recommendations with human oversight as areas for further discussion.
Key takeaways
- Move from principles to practice. Responsible AI policies are a starting point, not the endpoint. Governance needs to be examined in actual HR and Rewards processes, including job grading, payroll, recruitment, and performance.
- Define who controls the decision. Use the progression Inform → Challenge → Recommend → Decide to clarify AI’s role and determine who controls, challenges, and remains accountable for an outcome.
- Go beyond the “Human in the Loop” label. Human in the Loop was the dominant position among participants who responded, but meaningful oversight also requires attention to guardrails, principles, explanation, and accountability.
- Include embedded and vendor-provided AI. Governance should consider AI embedded in existing technology as well as externally provided AI capabilities. The meeting specifically raised the question of maintaining reliable outcomes in AI-enabled job matching supplied by vendors.
- Keep explainability and auditability visible. Concerns about the black box, bias, documentation, review, testing, and ultimately becoming Embedded & Auditable all point toward governance that can be examined in practice rather than existing only on paper.
AI generated by Copilot from designated meeting content only | Rumbold
