South Africa’s AI opportunity starts below the surface
Why your biggest AI gains will come from building the cloud, data and application foundations that turn promising ideas into business value
- By:
- Tarun Rao
There’s no shortage of AI experimentation in South Africa. Banks are applying it to fraud detection and customer service, retailers to hyperpersonalization and forecasting, and mining companies to predictive maintenance and operational safety.
But how successful will these projects be in the next two to three years?
To create the most value from AI, you need to line up three things: valuable proprietary data; high-volume decisions or repeatable workflows; and a clear economic case for using AI to make them faster or better.
This puts South Africa’s financial services industry in a particularly strong position. Banks and insurers have rich data, relatively mature digital environments, and obvious opportunities in fraud, credit, claims, operations and advisory.
Telecommunications has similar advantages, while mining and industrial companies can tie AI directly to asset uptime, safety, energy efficiency and production. Retailers, meanwhile, have opportunities wherever customer and transaction data can improve demand forecasting, inventory, pricing and service.
But can they turn promising use cases into something that works safely at scale every day?
The AI race won’t be won by collecting pilots
Let’s say two banks launch similar AI initiatives. The first gives employees a collection of new AI tools and runs successful proofs of concept in several departments. The demonstrations are impressive, but teams have to solve their own data access, security, integration and governance problems.
The second starts with a specific process — perhaps claims handling or customer onboarding — and analyzes what would have to change throughout that process to improve turnaround time by 30%. Before putting AI into production, the bank modernizes its key applications and data systems, and establishes reusable security and governance controls.
After a year, the first bank may have more AI projects, but the second has changed how work gets done.
An NTT DATA report, Cloud-led innovation in the era of AI: The new rules for driving value with cloud, states that only 14% of organizations consider themselves “cloud-evolved” — the most advanced when it comes to cloud adoption and impact — while 50% say the need to modernize applications and data platforms is holding back cloud-related innovation.
The research covered 2,335 decision-makers in 33 countries, including 75 in South Africa (the published findings cited here are global, not South Africa-specific), and the results show a clear divide between organizations that can put AI to work inside their core business processes and those that are still counting pilots.
Cloud, data and AI are now the same conversation
It no longer makes sense to discuss cloud transformation, data modernization and AI as separate programs.
AI needs access to trusted data. It needs applications and application programming interfaces (APIs) through which it can take useful action. It also needs scalable computing power, identity controls, security, governance and observability. In other words, it depends on much of the same foundation organizations have been building through cloud and application modernization.
The relationship also works in reverse. AI can accelerate cloud transformation by helping discover workloads, map dependencies, modernize code and streamline environments. But simply layering AI over an aging technology estate won’t get you very far.
Our research identifies data readiness and analytics as the leading source of dissatisfaction with efforts to build cloud-native AI applications. And 36% of organizations say agentic AI has intensified the need for clean, well-governed, AI-ready data architectures.
For South African businesses, the practical lesson is simple: Start with the outcome, then modernize the cloud, application and data foundations needed to achieve it.
Sovereignty doesn’t have to stop innovation
There is another complication, particularly when AI starts accessing and processing sensitive customer, citizen or operational data. Where should everything run?
It’s tempting to turn data sovereignty into a binary question about whether information is stored inside or outside South Africa. In practice, you need to think more broadly about control: Who controls the data, identities and encryption keys? Who can access them? Which jurisdictions apply? Can those controls be demonstrated and audited?
In addition, how do the Protection of Personal Information Act — which regulates cross-border data transfers — and the National Policy on Data and Cloud affect these decisions?
The answers will not be identical for every workload. A retailer’s product-recommendation engine and a bank’s AI system handling sensitive customer information don’t necessarily belong in the same environment.
While public cloud may be right for some workloads, private, sovereign or locally resident environments may suit others. Large organizations often operate across several of these, making consistent governance and integration more important than choosing one model for everything.
Globally, the direction is clear. According to our research, 99% of organizations expect private-cloud adoption to increase, while sovereign-cloud adoption is projected to rise 50% within two years. Integration with existing hybrid and multicloud environments is the leading concern around sovereign cloud.
The objective, then, is not to choose between innovation and control but to design control into the architecture so that you can scale sensitive AI without governance becoming an obstacle down the line. There’s strategic value in working with a partner like NTT DATA to make these workload-placement decisions in complex distributed estates.
The two essential investments I recommend
For the next 12 to 18 months, I would concentrate investment in two places:
1. Modernize your data and application core around a small number of high-value business domains
Pick two or three processes where better decisions or automation will materially affect revenue, cost, risk or customer experience. Next, fix the data quality, lineage, governance, integration and real-time access related to those processes. Modernize the applications where necessary. Then embed AI into the work itself.
2. Build a common cloud and AI platform rather than engineering every use case from scratch
Build the common plumbing once, then reuse it. Identity, security, cloud environments, cost controls, monitoring, automation and AI governance should not have to be reinvented for every new use case. This is important in a market where scarce specialist skills are better spent solving distinctive business problems than repeatedly rebuilding technical foundations.
Our research supports this direction. The proportion of organizations looking to use fully managed and automated cloud platforms delivered by strategic partners is set to rise from 11% to 32% in the next 12 to 18 months. Those partners provide the right skills, on demand, at a time when AI skills are ranked as the number-one cloud skills gap.
To make AI work, don’t take shortcuts
South Africa has pockets of sophisticated capability in banking, telecommunications, retail and industry, but it also has legacy technology, cost pressure and a scarcity of skills. That combination creates an interesting possibility: you don’t have to repeat every stage of earlier cloud transformations. AI can speed up modernization, while industry platforms shorten development cycles and managed platforms make specialist skills go further.
Even so, there’s no shortcut around the foundations. Cloud, data and AI are becoming parts of the same operating environment, and that’s where the next phase of the AI opportunity lies: not in having more experiments, but in becoming very good at turning the right ones into secure, repeatable and economically viable ways of working.
What to do next
Read more about NTT DATA’s Cloud services to see how we can help you build the cloud, data and application foundations you need to put AI to work at scale.
Author
Tarun Rao
Tarun Rao is Managing Director of Cloud and Security at NTT DATA South Africa. His focus is helping South African organizations move AI, cloud and data from promising pilots to secure, governed operations that deliver measurable returns. He has close to two decades of experience that spans global technology firms, management consulting and IT service management, combining a commercial view of where technology creates value with a practitioner’s understanding of how it is run at scale.