MTN's first AI target? Itself - Charles Molapisi — Charles Molapisi

MTN Group is targeting R30-billion in “value creation” from AI over the next three to five years, with roughly half expected to come from using AI to run the business more cheaply and the balance split between consumer and business applications.

Setting out the plan at the group’s capital markets day on Wednesday, chief technology and information officer Charles Molapisi said MTN would prioritise internal efficiency – the bet it is most confident about – while moving more cautiously on the riskier, capital-intensive push to sell AI infrastructure to other businesses.

AI at MTN is not one programme but a portfolio of three, each with a different risk profile and funding model:

AI Inside uses AI to run MTN itself – networks, energy, finance, HR, legal and fraud.
AI for B2C uses AI to sell more to consumers through hyper-personalised offers.
AI for B2B is the commercial play: renting out computing power, GPUs and edge capacity.

Read: MTN South Africa hunts up to R6-billion in savings

Each is measured differently. AI Inside is a margin lever – value is simply operating costs saved plus capital spending avoided. AI for B2C is gated by customer adoption and the cost to serve. AI for B2B is the biggest potential prize but the riskiest – capital-intensive and contestable, hinging on energy costs and “token economics”, the price of producing AI output.

“We will prioritise AI Inside, but we are executing B2C and B2B – we’re just a little more nuanced in terms of timing and capital allocation,” Molapisi said.

The group is going after its cost base first. The logic for this is in MTN’s income statement: network and IT account for around 55% of group operating expenditure and roughly 80% of capital expenditure. “It therefore follows that this is what we have to go after if we want to create value,” he said. The biggest single target is energy.

Several internal use cases are already live:

Energy optimisation manages power draw at base stations, with early deployments in the Western Cape; Molapisi argued the difference from past telco energy promises is scale, with AI able to interrogate vastly more parameters.
Autonomous networks use agents to triage incidents in South Africa and, via an approval process, tune some parameters automatically.
Fibre-cut sensing turns the fibre itself into a sensor that listens for micro-vibrations and flags digging before the spade goes in – aiming to collapse repair windows that can easily run to eight hours or more.
Smart capex forecasts the traffic and revenue profile of a locality six to 12 months out, so MTN can site mobile (AI RAN, or radio access network) and fixed (AI FAN, or fixed access network) builds with far more precision.
In the back office, legal contracts are being loaded into a single vault so AI can draft, amend and flag regulatory changes.

Selling to consumers

On the consumer side, the flagship is NBx 2.0, a “next-best-action” engine that targets individuals using richer data. It is live across six markets – South Africa, Nigeria, Ghana, Cameroon, Zambia and Uganda – reaching 44 million customers, and is scaling to all. Alongside it sit Telco GPT, an assistant for customers and call centre agents, and Zigi, a conversational tool that now works by text and voice in local languages.

Read: MTN to turn its African towers into an AI inference grid

The use case Molapisi returned to most was Sim registration in Nigeria. Until recently, more than 200 staff manually “eyeballed” incoming records, matching biometric data against the central database. MTN has replaced that with 13 AI agents using computer vision, now handling the bulk of processing autonomously and, it says, faster and more accurately. It is scaling to Cameroon, Côte d’Ivoire, Eswatini, Ghana and Zambia – with governance and “human in the loop” safeguards. A fourth live case, revenue assurance, scans 10 000-plus data points across billions of daily transactions to catch anomalies “humans cannot diagnose”.

To hold itself to account, MTN has built an “AI intensity formula” with a 2028 target: in its top six markets, 80% of AI’s bottom-line impact should come from AI Inside. That rests on 70% of the workforce trained, 40% “model and agentic penetration”, and 80% maturity in underlying technology and processes.

Access will be targeted to roles where value justifies cost.