Which workforce decisions should AI be allowed to make?
Set authority, evidence and challenge requirements for consequential decisions
PAWorld27 Zurich · 3–4 March 2027 · Zürich
About this session
A leadership team proposes one assurance process for both a routine HR information tool and a model that informs redeployment. In this worked leadership scenario, validation exposes missing experience data and uneven errors, and a human challenge changes a recommendation.
The two uses carry very different consequences. Leaders must decide where automation is acceptable, where AI can only advise, and where use should be deferred until the evidence or controls improve.
Learning outcomes
- Work through a decision-rights and assurance matrix: permitted use, minimum evidence, human authority, challenge route, monitoring and stop conditions.
- Explain which evidence or assumption changed the group's preferred choice.
- Name the accountable executive and the trigger for suspending use.
Strategic questions
Where is the boundary between automated action, decision support and deferred use, and which executive is accountable for the consequences and for handling challenges?
Context
As AI moves from analysis into recommendations about people, accountability has to be explicit. This Decision Room builds on the main-room sessions about a people analytics team's use of AI agents, a challenged AI recommendation and where automated analysis should stop.
