AI lessons from SA’s top digital newsmakers of 2025

Telkom sold its tower business for R6.75-billion and used the cash to strengthen its balance sheet. Vodacom spent R36-billion to take control of Safaricom. Cell C returned to the public markets, raising R2.7-billion at listing. These aren’t just “big corporate moves” from TechCentral’s South African Newsmakers of 2025—they’re signals about where South Africa’s digital economy is heading.

Here’s the stance I’ll take: AI in South Africa isn’t being “driven by chatbots.” It’s being driven by infrastructure, distribution, and the ability to turn customer data into decisions. The people on TechCentral’s list mattered in 2025 because they moved the pieces that determine how fast AI-powered e-commerce and digital services can scale—connectivity, content, fintech rails, and policy.

This post is part of our How AI Is Powering E-commerce and Digital Services in South Africa series, and the goal is practical: what these leadership stories mean for online retailers, marketplaces, subscription businesses, and digital service providers trying to win in 2026.

The real AI advantage: distribution beats “features”

AI adoption rises fastest where distribution is strong. In South Africa, distribution often means mobile networks, fibre reach, payments adoption, and content platforms—not just software.

TechCentral’s 2025 newsmakers highlight that the competitive edge is shifting toward companies that can do three things reliably:

1. Acquire customers at lower cost (because they already own a channel: SIMs, bundles, subscriber bases, merchant networks).
2. Learn from behaviour data (usage, viewing, buying, churn signals).
3. Act on those signals quickly (pricing, offers, network investment, fraud controls, retention flows).

For e-commerce and digital services, that’s the core AI loop: data → prediction → action → feedback. If your business doesn’t have the distribution or data to run that loop, your “AI strategy” will stay stuck in pilots.

What this means for South African e-commerce right now

As we head into the year-end/January retail cycle (returns, back-to-school spend, subscription downgrades), businesses that can predict churn, optimise fulfilment, and personalise offers will protect margin. But they’ll only do it at scale if they’re standing on solid digital rails—connectivity, payments, and identity.

Telkom’s play: why infrastructure choices shape AI outcomes

Telkom CEO Serame Taukobong made the list for turning Telkom into a more focused, data-led infrastructure player via “OneTelkom” integration—bringing mobile, fibre, and data centre pieces closer together. TechCentral reported Telkom’s 25% surge in EBITDA for the year ended 31 March 2025 and a return to dividends after a 62% jump in full-year earnings including Swiftnet proceeds.

Here’s the AI angle most people miss: integrated infrastructure reduces the friction between data collection and action.

When your network, fibre footprint, and compute capacity are aligned, you can:

• Improve real-time decisioning (who gets which offer, when, and at what price).
• Run more analytics closer to where the data is generated (edge and regional compute matters for latency and cost).
• Make experimentation cheaper (A/B tests, offer tests, segmentation updates).

Practical e-commerce takeaway: treat connectivity as a growth lever

If you sell online in South Africa, your conversion rate is tied to network reliability, speed, and the cost of data. That’s not abstract. It shows up as:

• abandoned carts during payment verification,
• slower product page loads,
• customer support contacts that could’ve been prevented,
• lower uptake of richer formats (video commerce, live shopping, interactive catalogues).

Retailers should negotiate partnerships (zero-rated flows, data bundles, loyalty tie-ins) like they negotiate payment fees—because it directly affects the AI loop that powers personalisation.

Vodacom’s consolidation: AI needs scale, and scale needs permission

Vodacom CEO Shameel Joosub is on the list for two major milestones: getting the long-contested Maziv fibre transaction across the line and moving to take control of Safaricom in a R36-billion deal.

The business logic is clear: AI thrives on scale. More customers and more transactions mean better models, better segmentation, and more resilient unit economics. But scale in telecoms and adjacent digital services also requires regulatory and structural permission.

Where AI shows up in telecom-led digital services

Vodacom’s push beyond connectivity into fintech and digital services matters for e-commerce because it shapes:

• identity and risk signals (SIM history, device reputation, location consistency),
• payment behaviour (wallet usage, top-ups, repayment patterns),
• offer delivery (bundles, personalised pricing, always-on channels).

If you run an online store, marketplace, or subscription service, you should assume telecoms will increasingly compete with you in adjacent categories (payments, merchant services, ad inventory, and even shopping experiences). That’s not a threat by default, but it does change what a “good partnership” looks like.

Practical e-commerce takeaway: build for multiple rails

I’ve found that South African businesses get stuck when they design their checkout, verification, and messaging around a single provider. AI-powered growth needs resilience.

A simple 2026-ready checklist:

• Two payment options minimum (card + instant EFT / wallet / pay-by-bank).
• Two messaging channels for key flows (SMS/WhatsApp/email).
• A fallback identity path when OTP delivery fails.
• Fraud controls that don’t rely on one signal source.

That’s not glamorous, but it’s how you keep your AI models learning instead of starving them with broken funnels.

Cell C and the MVNO surge: the quiet data revolution

Cell C CEO Jorge Mendes made the list after Cell C’s JSE listing in November, raising R2.7-billion at R26.50/share via a stake sale by The Prepaid Company (Blu Label). The turnaround leaned into an asset-light model and, importantly, deepened Cell C’s role as an MVNO host—including fast-growing Capitec Connect.

This matters for AI because MVNOs are often built around tight segments (bank customers, retail loyalty bases, youth bundles). Segmentation is where AI actually pays.

Why MVNOs are an AI-friendly business model

MVNOs can:

• use first-party data from a bank/retailer ecosystem (spend categories, salary timing, device finance),
• run more targeted lifecycle offers (top-up nudges, roaming packs, payday bundles),
• tie network value directly to loyalty and rewards.

For e-commerce, MVNO growth signals something bigger: customer acquisition is increasingly bundle-based. Your future “traffic source” may be a partner’s subscriber base with AI-driven targeting, not a generic social ad.

Practical e-commerce takeaway: start thinking in bundles, not campaigns

If you sell products or subscriptions, test bundle partnerships that combine:

• data + delivery (discounted data to browse and buy),
• loyalty + credit (buy-now-pay-later aligned to customer risk),
• service + content (value adds that reduce churn).

AI makes these bundles profitable by predicting who will respond and who will churn. Without AI, bundles can become expensive guesswork.

Canal+ and MultiChoice: content platforms are AI platforms

Canal+ CEO Maxime Saada made the list for driving the takeover of MultiChoice, reshaping the continent’s broadcasting map. The strategic bet is scale across Africa while the market shifts from linear TV to streaming, with escalating content costs and fierce competition.

For digital services, the key point is direct: recommendation engines, ad targeting, and retention modelling are AI problems. Streaming platforms are effectively training South African consumers to expect:

• personalised discovery,
• frictionless payments,
• predictable monthly value.

That expectation spills into e-commerce. Your shoppers now compare your product discovery to their streaming homepage, whether that’s fair or not.

Practical e-commerce takeaway: improve discovery before you spend more on ads

Most businesses throw budget at acquisition because discovery is weak. Fixing discovery is often cheaper than buying more traffic.

Three high-impact AI use cases for South African online retail:

1. Search that understands intent (synonyms, misspellings, local naming).
2. Recommendations that respect constraints (size, delivery area, affordability).
3. Retention triggers based on behaviour (browse → abandon → return patterns).

If you can’t do “full AI personalisation” yet, start with rules informed by data: top products per segment, seasonal bundles, and smart reordering prompts.

Nkosana Makate and the policy layer: trust, IP, and incentives

Nkosana Makate topped the list after a 17-year legal battle around the “please call me” concept, reportedly settling out of court in November (with further litigation risk emerging). This story isn’t about AI directly, but it’s central to what happens next in the digital economy: people build what they believe they can own.

AI in e-commerce and digital services raises the same questions—just at speed:

• Who owns training data?
• Who owns prompts and outputs?
• What happens when an employee creates a valuable workflow?

If incentives and IP are unclear, innovation slows or gets hidden.

Practical e-commerce takeaway: set AI governance before it becomes a fight

A workable, non-bureaucratic governance pack for SMEs and mid-market teams:

• A one-page AI usage policy (what data can/can’t go into tools).
• Clear rules for customer data and consent (Popia-aligned).
• Clauses on IP ownership for internal AI workflows and content.
• A review process for high-risk outputs (pricing, credit decisions, medical claims).

The 2026 playbook: what to do next (even if you’re not “AI mature”)

AI strategy dies when it’s treated as a software shopping list. The newsmakers of 2025 show that winners build foundations—distribution, data, and operating discipline.

Here’s what I’d do in the next 60 days if I ran an e-commerce or digital services business in South Africa:

1. Map your data pipeline: where behavioural data is captured, where it gets stuck, and what you can actually use.
2. Pick one metric that matters: conversion rate, repeat purchase rate, churn, or fraud loss.
3. Ship one automation tied to that metric: cart recovery, smart offers, fraud scoring rules, or customer support triage.
4. Secure two partners that expand distribution: a payments partner and a channel partner (telco, bank, loyalty platform, marketplace).
5. Measure weekly and keep model ambition modest until your data quality improves.

AI becomes profitable when it changes decisions, not when it generates content.

The bigger question heading into 2026 is simple: will your business be a destination, or will it be a feature inside someone else’s bundle? The moves made by South Africa’s top tech leaders in 2025 suggest the bundle economy is accelerating—and AI is the engine that makes it financially viable.

If you want help prioritising AI use cases for your e-commerce stack or digital service, start with the fundamentals: clean data, resilient funnels, and one growth loop you can measure end to end.