Insights

Practical thinking on AI in Africa

Governance, security, architecture, policy and sovereign AI, written for the people who have to make it work.

Your AI risk assessments describe systems that no longer exist

Every AI approval rests on an assessment of one model version, one vendor contract and the rules in force that day. As African banks, telcos, insurers and public agencies scale their AI portfolios, here is how to keep risk assessments current without growing the team.

Policy5 min read

Preparing for risk-based AI regulation in Africa: a practical checklist

Risk-based AI governance is taking shape across the continent. This checklist covers the groundwork that holds regardless of the final wording of any law, from AI inventories to data protection under POPIA, NDPA and Kenya's DPA.

Governance4 min read

Fenced AI: why a working model is only 10% of the job

A capable model is the starting point, not the finish line. The Fenced AI rings of Capability, Architecture, Governance, Security, Content Safety and Infrastructure Fit cover the rest of the work.

Architecture4 min read

Why Africa can't copy-paste Western AI

Most AI products assume stable power, cheap bandwidth, English-first users and nearby cloud regions. Here is how to design AI architecture for the conditions African organisations actually operate in.