For most of the past decade, digital government has meant digitising what already existed: moving paper forms online, replacing queues with portals, converting manual registers into databases. While useful, this largely automated processes instead of changing them in any meaningful way.
By Lance Williams, public sector lead at SAP Africa
The arrival of AI agents is starting to shift this. Governments are now moving from digitising administration to deploying AI agents that can plan, decide and act across multi-step workflows, processing applications, managing records, flagging exceptions, and responding to citizen queries with far less manual intervention.
Gartner projects that at least 80% of governments will be using AI agents to automate this kind of routine decision-making by 2028.
From admin to outcomes
This shift is being framed as agentic AI moving public administration from task automation to outcome orchestration. Importantly, this is viewed as a model of augmentation rather than replacement, with AI taking on the coordination and case-processing load, freeing public servants to focus on the complex cases, policy interpretation and direct citizen engagement that genuinely require human judgment.
In several countries throughout the continent, evidence of government-led AI innovation is emerging. South Africa’s Revenue Service is pursuing AI-driven “digital twinning” to replicate human decision-logic in tax processing, turning resource-intensive manual work into consistent, scalable digital workflows.
In Rwanda, AI-assisted medical analysis has reportedly cut diagnostic waiting times by more than 60% in rural hospitals since a 2024 pilot. Uganda’s newly launched GovAgenti platform – built explicitly as a transparency tool – lets citizens track public project delivery in real time across WhatsApp, USSD and 16 local languages, while Kenya’s eCitizen platform now carries more than 22,000 government services and 12 million active users.
Governance the key to trust
While every serious conversation about algorithmic decision-making in government eventually reaches the same reassurance (“Don’t worry, a human is in the loop”), human presence is not the same thing as governance.
A reviewer can technically satisfy a human-in-the-loop requirement while lacking the authority, the information or context, or simply the time to meaningfully challenge what an AI system has produced. Research examining comparable requirements elsewhere has found little evidence that a human checkpoint alone improves outcomes; without the right infrastructure behind it, the safeguard becomes symbolic.
What protects citizens is governance built before deployment, not review bolted on afterwards: procurement standards that specify explainability up front, defined escalation paths for contested decisions, complete audit trails, and named officials with the real authority to override or suspend a system.
SAP’s own approach reflects this philosophy. Rather than viewing AI as an independent decision-maker, it is designed to operate within governed business processes where permissions, auditability, human oversight and responsible AI principles are built into the underlying platform. AI assistants and agents can help automate routine work, surface relevant information and coordinate complex workflows, while ensuring that accountability remains firmly with authorised public officials.
That balance between automation and oversight will become increasingly important as governments move beyond isolated AI pilots towards broader transformation.
Citizen experience is the ultimate measure
Ultimately, citizens are unlikely to judge governments on how much artificial intelligence they deploy. They will judge them on whether renewing an identity document becomes easier, whether benefits reach eligible recipients more quickly, whether procurement becomes more transparent, and whether services become more reliable and consistent.
The governments that lead the next phase of digital transformation will not necessarily be those deploying the most AI. They will be those that use it most responsibly: reducing delays, improving consistency, strengthening accountability and allowing public servants to spend less time navigating administration and more time serving citizens.
That is the promise of the autonomous public service: not government run by machines, but government empowered by trusted intelligence, with people remaining firmly responsible for the decisions that matter most.