AI agents could change face of service delivery

Artificial intelligence (AI) is moving beyond chatbots and digital forms, with some governments exploring how technology can be used to coordinate tasks, process applications and support public services. Experts warn, however, that as machines take on more responsibility, accountability cannot also be automated. For the past few decades, many governments’ digital footprints have largely been about transferring services online; that is, transforming paper forms into digital applications, replacing physical queues with online portals and turning manual registers into databases. Many of the changes have made some government processes more accessible but they have not necessarily changed the way the system works. Now, the rapid development of AI could usher in a different phase of digital transformation. AI agents are being developed to do more than answer questions. They can potentially plan tasks, make recommendations, process information and coordinate multiple steps in workflows with limited human intervention. This could mean AI systems helping to process applications, manage records, identify exceptions, respond to citizen queries and route cases to the appropriate officials. Gartner has projected that at least 80% of governments will be using AI agents to automate routine decision-making by 2028. The shift moves digital transformation away from simply automating individual administrative tasks towards coordinating entire processes. The concept is being described as “agentic AI” — systems capable of carrying out multistep tasks rather than simply responding to individual prompts. The idea is not necessarily to replace public servants. Instead, AI could take on some of the coordination and case-processing burden, while officials remain responsible for interpreting policy, dealing with exceptions and engaging directly with citizens. The South African Revenue Service, for example, has been exploring AI-driven “digital twinning” approaches designed to replicate human decision logic in tax processing. The potential benefits are clear: faster processing, greater consistency and fewer administrative bottlenecks. But there is another question that becomes important as AI takes on more responsibility: who is accountable when the system gets it wrong? For AI to be trusted in public services, accountability needs to be built into the system. This includes clear procurement standards, defined escalation procedures, comprehensive audit trails and named officials who have the authority to override or suspend an AI system. Lance Williams, the public sector lead at SAP Africa, says AI should be viewed as a tool for strengthening public administration rather than removing accountability. “The real opportunity is to use AI to remove administrative friction, help public servants make better-informed decisions and ultimately improve the experience of citizens.” Williams says governance and accountability need to be considered. “AI agents should operate within clearly governed processes, with appropriate permissions, auditability and human oversight ... but accountability for decisions that affect citizens must remain with authorised public officials.” He says the measure of success will be whether it helps government deliver services that are faster, more consistent, transparent and accessible. Putting an AI agent on top of a broken or poorly designed administrative system will not necessarily produce better outcomes. Governments will need to examine their processes, data, policies and accountability structures before deciding where AI can add value.