The author of this working paper argues that there is no substitute for sound, evidence-based policy development to support a human-centric approach to social policy and government to service delivery. To enhance the benefits from AI while minimising the adverse risks, social security administration needs to understand better the scope and depth of the risks...