Delegates at a recent AI summit in Johannesburg urged a shift from hype to hard work: Africa can become an AI creator and exporter, not merely a consumer, provided governments and industry first tackle the continent’s patchy digitisation, ageing infrastructure and security challenges.
Foundations before features
Senior executives from the energy, banking, security, work and investment sectors used the summit to stress that practical building blocks must be in place before advanced AI systems are rolled out. The consensus was that useful, trustworthy AI depends on five pillars: local data, relevant use cases, skilled people, resilient infrastructure and security-by-design.
Schneider Electric’s Steven Santini, Vice President of Secure Power for Sub‑Saharan Africa, told the summit that technology cannot produce value if physical equipment produces no data. He noted, according to reporting from the event, that in some areas only one traffic light in five is functional — a simple example of how analogue assets leave algorithms with nothing to act on.
“We need to take a step back and get the fundamentals right before we rush to meet the demand for AI, otherwise we’re just setting ourselves up for failure,” the report quoted Santini as saying.
Speakers emphasised that digitisation must come first and that AI should accelerate existing capabilities rather than replace them overnight. Without reliable electricity, stable communications and data collection from the edge, advanced models and so‑called agentic AI risk being ineffective or introducing new risks.
Agentic AI is an infrastructure and security problem
Cisco’s Nabeel Rajab, identified at the conference as a Technical Solutions Architect in South Africa, framed the move from chatbots to agents as a structural shift. He said the transition changes where and how computing and security need to be designed, because agents can run chains of tools and workloads that place heavy demands on networks and on threat defences.
- Local data: models trained on Africa‑specific data to avoid imported biases and ensure relevance.
- Practical use cases: solutions that address real sector problems in energy, banking, security and labour markets.
- Capable people: skills development across technical and operational roles.
- Resilient infrastructure: reliable power and communications to support data collection and compute.
- Security‑by‑design: protecting systems from the outset, especially where legacy networks persist.
The caution voiced at the summit reflects the continent’s uneven infrastructure. Many assets remain analogue; power grids and communications networks vary in reliability; and connectivity gaps mean large populations are still offline or poorly served. These conditions make it harder to collect representative data and to sustain continuous, secure AI services.
| Challenge | Implication for AI |
|---|---|
| Analogue infrastructure | Insufficient telemetry for models to act on |
| Ageing power systems | Intermittent compute and unreliable services |
| Uneven connectivity | Limits on data collection and distributed workloads |
| Legacy network security | Higher risk when deploying agentic systems |
What this means for South Africa
For South Africa, the summit’s messages map onto familiar policy debates: investment in grid resilience, broadband and edge digitisation; skills programmes that go beyond headline AI courses; and tighter governance of data and security. Industry leaders at the event argued that without these fundamentals, attempts to import finished AI products will underdeliver and could amplify existing inequalities.
The conference tone was cautiously optimistic: the continent has potential to design and export AI if it focuses on sovereignty, local relevance and trust rather than chasing external narratives of rapid disruption. That requires patience and funding for basic systems — not just for model licensing and cloud credits.
As South Africa and its neighbours consider AI strategies, the summit’s takeaway is clear: get the basics right, build from local needs and secure systems by design, and only then scale intelligent agents that can truly add value.
