African government tech leaders flag data and skills gaps hampering public-sector AI governance

TechTrends Kenya reports on a panel about AI in government held at WSO2Con Africa 2026 in Nairobi, held September 22 to 24 and themed "Building the Agentic Enterprise". Speakers came from Uganda's National Information Technology Authority, the Zambia Revenue Authority, Oman's Social Protection Fund and WSO2. They argued that the hard work in public-sector AI begins before any model is deployed.

Harrison Chapu of the Zambia Revenue Authority said projects should start by identifying the problem and checking whether AI is appropriate, since conventional automation may suffice. Tonny Bbosa of Uganda's NITA-U said his team defines the intended outcome and looks for measurable gains, such as fewer physical visits by citizens or shorter processing times. The panel treated data as the main constraint. Chapu said the authority's data is segmented and tied to legacy systems, which raises questions about integrity. He suggested imperfect data can be tolerated for advisory uses with a human deciding, but not for agents making consequential decisions alone. Hamda Al Issaei said mismatched identifiers, missing standards and weak integration all carry through into AI results.

The panel also cited a diabetic retinopathy screening project across more than 35 healthcare institutions as a practical example, while noting challenges with data preparation, integration and security. Chapu argued civil servants need training that separates automation from AI and varies by role. Looking ahead, speakers described more proactive government, with fewer office visits and services that anticipate needs. Chapu's example of a revenue authority spotting activity before a taxpayer reports it was presented as a projection, not an existing service. The article concludes that AI ambition should match data quality, integration maturity and available skills.