AI-powered digital services: SA’s 2025 power players

South Africa’s digital economy didn’t grow in 2025 because everyone “discovered AI”. It grew because a handful of leaders made unglamorous moves that changed the plumbing: fibre consolidation, mobile price pressure, MVNO expansion, media scale plays, and even court fights that forced big corporates to take IP claims seriously.

If you run an online store, a marketplace, a subscription service, or a fintech product, this matters. AI in e-commerce and digital services only works when connectivity is affordable, data is accessible, payments are frictionless, and platforms can scale. The newsmakers TechCentral highlighted this year sit right on those levers.

This post is part of our series on how AI is powering e-commerce and digital services in South Africa. The angle here is practical: what these leadership stories tell you about where to place your bets in 2026—and what to fix in your own business before you spend another rand on “AI initiatives”.

The real AI bottleneck in South Africa: infrastructure, not ideas

Answer first: For most South African businesses, AI success is constrained more by connectivity costs, latency, data quality, and customer trust than by model choice.

The market loves talking about chatbots and product recommendations. But the businesses that win with AI tend to get the basics right first:

• Reliable, reasonably priced data access so customers can actually use your service (and so your team can ship updates quickly).
• Distribution at scale (mobile networks, fibre, platforms) because AI improvements compound when you have volume.
• Clean, governed data because AI trained on messy inputs becomes an expensive confusion machine.

TechCentral’s list is heavy on telecoms and media for a reason. When fibre consolidates, when prepaid competition intensifies, when MVNOs scale—your customer acquisition cost and retention dynamics shift.

A seasonal note: we’re writing this just after the 2025 holiday surge and the Black Friday/Cyber Monday cycle. That period exposes every weakness in the stack: delivery promises, payment failures, customer support queues, fraud spikes. AI helps—but only if your underlying systems don’t fall apart under load.

Telecom leadership is reshaping what “digital-first” costs

Answer first: The 2025 telecom newsmakers influenced the unit economics of digital services—data pricing, fibre access, and wholesale models—directly affecting AI-driven growth.

Telkom’s “OneTelkom” lesson: unify systems before you automate

Serame Taukobong’s Telkom story isn’t “about AI” on the surface. It’s about integration: mobile, fibre, and data centres working as one business. That matters because AI projects hate silos.

Telkom’s reported performance improvements in 2025 (including a 25% EBITDA surge for the year ended 31 March 2025 and reinstated dividends after a 62% full-year earnings jump including Swiftnet proceeds) point to something operators and retailers both learn the hard way:

If your organisation can’t share data across teams, your AI will turn into a set of disconnected pilots.

If you’re in e-commerce, copy the principle:

• One product catalogue (not five versions across channels)
• One customer profile (not fragmented identities across web, app, WhatsApp, and stores)
• One measurement layer (so you can attribute revenue properly)

Practical AI takeaway: Before you roll out advanced personalisation, build a “single view” foundation: identity resolution, event tracking, and product data hygiene.

Vodacom, Maziv and fibre consolidation: speed and scale change conversion

Shameel Joosub’s year was defined by finally getting the Maziv fibre transaction across the line (after years of regulatory friction) and pushing regional scale via Safaricom.

For AI-powered e-commerce and digital services, fibre consolidation tends to produce two downstream effects:

1. Better network economics over time (more coordinated capex, broader coverage, stronger wholesale positioning)
2. More consistent experience for consumers who shop, stream, and pay online

This is where AI becomes a multiplier. When your customer’s connection is stable, you can:

• Serve richer product imagery/video without tanking load times
• Use real-time fraud scoring during payment
• Offer proactive support (predicting churn or delivery issues)

My stance: South African businesses should treat network improvements as an opportunity to raise their service bar—not just to run heavier pages. Faster connectivity should translate into fewer steps to checkout, faster support resolution, and smarter post-purchase comms.

Cell C and the MVNO model: distribution isn’t optional

Jorge Mendes’ Cell C turnaround and JSE listing highlights the growing weight of MVNOs in SA (including high-profile banking-led MVNOs).

That matters because AI-powered digital services are increasingly bundled:

• Banking apps + mobile bundles
• Retail loyalty + mobile data rewards
• Subscription services + device financing

MVNO growth pushes customer expectations: more self-service, more real-time notifications, fewer branch/store interactions. AI plays nicely here, especially in:

• Automated onboarding and KYC support
• Personalised data bundles and rewards
• Support deflection via well-designed messaging flows (not “sorry, I didn’t understand that” loops)

Practical AI takeaway: If you operate in fintech, retail, or subscriptions, map your distribution partners and ask a blunt question: Are you building features that partners can actually bundle and sell? AI features that don’t fit a bundle often die in a backlog.

Media scale and streaming wars are training customers for AI-driven experiences

Answer first: The Canal+/MultiChoice deal reflects a broader shift: customers now expect on-demand, personalised, multi-device experiences—the same expectations they bring to shopping and digital services.

Maxime Saada’s presence on a “South African” list makes sense because the Canal+ takeover of MultiChoice reshapes the continent’s content distribution dynamics. But the bigger spillover is behavioural.

Streaming competition has normalised:

• Personalised home screens
• “Continue watching” journeys
• Smart recommendations
• Subscription churn management

That expectation leaks into commerce. Customers now assume:

• Your store will remember their size, shade, or preferred delivery slot
• Support will be instant and contextual
• Pricing and promos will be relevant (or they’ll ignore them)

Here’s what works in practice for SA e-commerce teams trying to match that standard without overbuilding:

1. Start with lifecycle personalisation (welcome, replenishment, win-back) before algorithmic “you may also like”.
2. Use AI for catalog enrichment (better titles, attributes, translations) because it improves search and SEO at the same time.
3. Prioritise customer support triage (intent detection, routing, summarisation) because it reduces resolution time during peak periods.

Personalisation that breaks trust is worse than no personalisation.

If you’re using customer data, be explicit: what you collect, why, and how it benefits the customer. Popia-aligned clarity is not paperwork—it’s conversion.

IP, policy, and trust: the “soft” factors that decide AI adoption

Answer first: South Africa’s AI opportunity is shaped by trust—how policy encourages investment, and how IP disputes signal whether innovators will be rewarded.

The Makate lesson: innovation needs enforceable outcomes

Nkosana Makate’s long-running “please call me” battle is a reminder that digital value isn’t only created by code; it’s created by ideas, execution, and the willingness to fight for recognition.

For AI-driven businesses, this translates into two concrete behaviours:

• Contract properly with creators and developers. If you commission datasets, prompts, training workflows, or customer conversation designs, document ownership.
• Treat internal innovation as an asset. Track experiments, record decisions, and keep evidence. When value shows up later, you’ll need the history.

AI increases the number of “micro-inventions” inside a company: new automations, new decision rules, new data pipelines. Without governance, you get confusion about who owns what and why it matters.

Policy momentum: investment and execution matter more than slogans

TechCentral’s notable mention of communications minister Solly Malatsi points to a reality: policy can either attract capital into networks and platforms or stall it.

From a business perspective, you can’t control policy—but you can prepare for it:

• Build services that degrade gracefully when connectivity is weak
• Diversify acquisition channels (search, social, marketplaces, WhatsApp)
• Avoid vendor lock-in that makes compliance or hosting changes painful

My take: The companies that win in 2026 will be the ones that assume volatility—then design for it. That includes load-shedding knock-ons, network congestion, fraud waves, and shifting platform rules.

What to do next: an AI roadmap that fits South African realities

Answer first: The fastest path to ROI is to apply AI to revenue-critical workflows—catalog, conversion, retention, and support—while tightening data governance.

If you’re generating leads or trying to scale growth in 2026, use this checklist to prioritise.

1) Fix data foundations before model shopping

• Define 15–30 core events you’ll track (view item, add to cart, checkout start, payment success, delivery failure)
• Standardise product attributes (brand, size, colour, compatibility, warranty)
• Create a single customer ID strategy across channels

2) Use AI where it saves time and improves customer experience

High-impact, low-regret use cases for SA e-commerce and digital services:

1. Customer support summarisation (shorter handling time, better handoffs)
2. Fraud detection and risk scoring (fewer chargebacks, fewer false declines)
3. Catalog enrichment and SEO copy (better search results, higher organic traffic)
4. Lifecycle messaging automation (abandoned cart, back-in-stock, renewal reminders)

3) Measure what matters with three numbers

Pick three metrics you review weekly:

• Conversion rate (by device and channel)
• Repeat purchase rate (or renewal rate)
• Cost per resolved support case (or average handling time)

If AI doesn’t move at least one of those in 60–90 days, pause and rework the inputs.

Where this is heading in 2026

South Africa’s 2025 newsmakers show the pattern: scale and integration beat noise. Telecom consolidation, MVNO growth, platform bundling, and better policy execution will keep pulling more commerce and services into digital channels.

For teams building AI-powered e-commerce and digital services in South Africa, the play is straightforward: get your data in order, design for local constraints, and apply AI to the workflows that directly affect revenue and trust.

If you had to bet on one area for 2026, I’d bet on AI-assisted customer operations—support, fraud, payments, and delivery communications—because that’s where South African consumers feel pain most sharply, especially during peak season. What would you fix first: product discovery, checkout friction, or post-purchase support?