AI adoption in South Africa is not a public sector project, it is an SME imperative

In October 2024 the Department of Communications and Digital Technologies published South Africa's National Artificial Intelligence Policy Framework, the first formal step towards a national AI policy. It names AI a general-purpose technology, on the order of electricity or the internet, and sets out how the country intends to build talent, digital infrastructure, research capability and public sector implementation around it.

Read at speed, it looks like a government document about government. Read properly, it is a description of the market conditions every South African business is about to trade in. The framework's own diagnosis is blunt: technological advancement is happening globally whether or not we participate, and the country must adopt these innovations to stay competitive and relevant. That sentence does not stop at the doors of state-owned enterprises.

"National policy sets the weather. It does not fly the plane. The businesses that read this framework as somebody else's homework will be the ones competing on price against companies that have already rebuilt how they work." - Alexandra Zoë Varenzakis

Where the real economy sits

Small and medium enterprises carry the majority of South African employment and the majority of its economic texture, the suppliers, the service firms, the manufacturers, the agencies, the traders. Public sector implementation is one of the framework's nine strategic pillars, but the economic transformation it hopes for, reduced unemployment and a dynamic competitive environment, cannot be delivered by the public sector alone. It is delivered by private firms deciding, individually, to operate differently.

The framework is equally clear about what stands in the way: a persistent digital divide, historical inequities, institutional inertia and regulatory frameworks not built for this pace. Institutional inertia is not only a government condition. It is the owner-managed business running on the same processes it used in 2015 because those processes still, technically, work.

Digitalisation is the precondition, not the project

There is no meaningful AI adoption on top of undigitalised operations. If quotes live in a WhatsApp thread, stock lives in someone's head and client history lives in an inbox, there is nothing for intelligence to act on. Digitalising strategy, operations, finance, sales and service is the groundwork, and it is where most of the return actually comes from before a single model is deployed.

• —Strategy, decisions made against live data rather than last quarter's impression
• —Operations, workflows that are recorded, measurable and therefore improvable
• —Sales and service, response times and follow-up that no longer depend on who is in the office
• —Finance, cash, margin and pricing visible weekly rather than at year end
• —Knowledge, what the business knows held by the business, not only by its longest-serving staff

"You cannot automate a process you have never written down. Digitalisation is the discipline of making your business legible to itself, AI is what you do afterwards." - Alexandra Zoë Varenzakis

Sustainability, jobs and upskilling

The fear attached to this conversation is job loss. The framework's stated position, and ours, is the opposite: AI applications should augment human decision-making rather than replace it, and the economic case rests on enhancing productivity, creating new industries and fostering innovation. In practice, in a mid-sized South African firm, that shows up as a business that survives its next downturn, keeps its people, and moves those people into higher-value work.

Upskilling is the mechanism. Talent development is the framework's first strategic pillar for a reason, a workforce fluent in these tools is the difference between a business that adopts and a business that buys software nobody uses. The firms that invest in that fluency create roles that did not previously exist inside them: data-literate operators, automation owners, people who can brief and supervise a system rather than merely execute a task.

New revenue, new industries, new markets

The commercial upside is not only efficiency. When a business becomes digital in how it operates, expansion becomes materially cheaper. Productised services, subscription and retainer models, data-derived offerings and licensable know-how all become viable at a scale that previously required headcount the business could not afford.

• —Alternative revenue streams built on capability the business already has
• —Adjacent industries entered without rebuilding the operating model from scratch
• —Internationalisation, serving clients beyond South Africa's borders with the same core team
• —Partnerships and networks that only open once your operations can be integrated with

"Africa to the world is not a slogan. It is an operational question: can your business serve a client in another market on Monday morning without adding a single person? For a digitalised firm the answer is yes." - Alexandra Zoë Varenzakis

AI and human co-creation

The framework's insistence on human-centred AI (transparency, explainability, fairness, professional responsibility) is not a compliance footnote. It is a description of what actually works commercially. Systems deployed without human judgement produce confident, plausible, wrong output at scale. Human judgement without these systems produces good decisions too slowly to matter.

Co-creation is the working model: the machine handles pattern, volume and recall; the person supplies context, ethics, taste and accountability. Every engagement we run at Venture A is designed around that pairing, because it is the only version of AI adoption that holds up over years rather than quarters.

"AI does not replace judgement, it raises the cost of not having any. The firms that win here are the ones where people and systems build together, and a human still signs the decision." - Alexandra Zoë Varenzakis

What to do this quarter

• —Diagnose, map where your business is genuinely digital and where it is still manual and undocumented
• —Strategy, decide which two processes, if digitalised, change your competitive position most
• —Build, implement those two properly, with your people trained to run them
• —Iterate, measure, refine, then extend into revenue, new markets or new industries

The national policy will take its course. Your competitive position will not wait for it.