AI Governance Must Give Affected People Real Decision Rights

An institution can consult widely and still decide alone. It can invite teachers to a demonstration after procurement, ask learners to complete a satisfaction survey after assessment has changed, brief workers after automation has redistributed their tasks, or request community feedback after an AI system has already absorbed their data. Such activities create records of participation without necessarily sharing knowledge, influence or remedy.

Artificial intelligence intensifies this problem because systems can change faster than ordinary governance cycles. They may alter what counts as competent work, whose knowledge becomes visible, how people are classified and where responsibility is located. Affected people therefore need more than awareness. They need the capability and authority to shape purpose, implementation, evidence and correction.

The governing principle is simple: no consequential AI deployment should be considered participatory merely because people were present. Institutions should be able to show what participants knew, when they were engaged, what they could change, which interests were missing, how disagreements were recorded and where an affected person can still challenge the result.

On 4 September, South Africa's National Economic Development and Labour Council convened its thirty-first annual summit under the theme of inclusive growth and decent work in the age of AI. NEDLAC brings government, labour, business and community constituencies into national social dialogue. South African public reporting records the Employment and Labour Minister's call for…