Risk Management and Human Oversight in AI

Responsible implementation of AI in people practice sits at the heart of AC 3.3, and this video addresses it through the lens of risk management and human oversight. It covers the distinct categories of risk that AI systems introduce into HR contexts — including bias and discrimination, privacy violations, accuracy failures, transparency limitations, legal non-compliance with data protection and equality legislation, and dependency risks that can erode professional judgement over time. The video draws on concrete people practice settings, with recruitment screening used as an anchor example of where these risks are most acute.

A significant portion of the video is devoted to the principle of proportionality in AI governance — the idea that the level of scrutiny applied to an AI application should correspond to the severity of its potential impact on individuals. The video distinguishes high-risk use cases from lower-risk ones, and examines what characteristics place an application in each category. Formal risk assessment before implementation, acceptable use policies, data protection controls around sensitive information, and phased or pilot approaches to introducing AI tools all feature within this framework. Accountability structures — including the assignment of clear roles and responsibilities for AI use — are also addressed.

The concept of "human in the loop" receives detailed treatment, with the video examining what meaningful human oversight actually requires in practice, as distinct from a purely nominal human presence in an AI-assisted decision process. It connects this principle to the broader ethical and legal context in which HR teams operate, and touches on the training and capability building needed for people professionals to interrogate AI outputs, recognise when to override recommendations, and maintain accountability for decisions in areas such as recruitment and learning and development.