Telecom policy execution shapes AI e-commerce success in SA. Here’s what Malatsi’s directive and EEIP oversight mean for digital services—and what to do now.
Telecom policy execution: the AI e-commerce bottleneck
Black Friday 2025 showed how quickly South Africans shift spend online when the offer is right—and how quickly the experience falls apart when networks, pricing rules, and compliance uncertainty create friction. AI can personalise product pages, spot fraud in milliseconds, and run customer support at scale. But none of that matters if the underlying connectivity and regulatory rails stay unpredictable.
That’s why the telecoms industry’s reaction to communications minister Solly Malatsi’s recent policy directive to Icasa (focused on BEE and equity equivalent investment programmes, or EEIPs) is more than a political spat. It’s a signal to every online retailer, marketplace, fintech, and subscription service: the next phase of AI-powered e-commerce in South Africa depends on policy execution—not policy headlines.
In this post (part of the How AI Is Powering E-commerce and Digital Services in South Africa series), I’ll translate what’s happening in telecom policy into practical implications for digital operators, and what you can do now to reduce risk while still shipping AI features.
Why telecom policy is an AI issue (not just a telco issue)
Answer first: AI in e-commerce is only as good as the network, spectrum economics, and regulatory certainty supporting it.
Most teams treat telecom regulation as “infrastructure noise” in the background. That’s a mistake. In South Africa, connectivity isn’t merely a utility cost; it’s a strategic dependency that shapes whether AI initiatives succeed or stall.
Here’s the direct line from telecom policy to AI performance in digital services:
- Latency and reliability drive conversion. If your AI-driven personalisation makes pages heavier, any drop in network quality can wipe out the uplift.
- Data flows power machine learning. Product recommendations, fraud models, and churn prediction depend on continuous event streams. Flaky connectivity means missing signals.
- Compliance affects go-to-market speed. Regulatory uncertainty changes investment decisions—network rollouts, data centre choices, pricing, and partnerships.
Malatsi’s directive is rooted in transformation policy and market entry rules, but the telecoms industry’s warning—execution matters more than intent—is what should grab e-commerce leaders.
What the directive debate tells us about “execution risk”
Answer first: If oversight and enforcement are unclear, the market gets uneven rules, slower investment, and more legal/operational risk for everyone building on the network.
The Association for Comms & Technology (ACT), representing major operators (including Vodacom, MTN, Telkom, Liquid Intelligent Technologies, Cell C and Rain), broadly welcomed the directive’s direction while raising a pointed concern: EEIPs often lack transparency and credible monitoring.
ACT CEO Nomvuyiso Batyi’s argument is straightforward:
- The sector can live with new policy direction.
- The real danger is how Icasa implements it, and whether enforcement becomes inconsistent.
That “how” question matters because the directive touches competitive parity—especially when new entrants use EEIPs rather than traditional ownership structures. From an e-commerce perspective, inconsistent enforcement can lead to:
1) Unpredictable connectivity economics
If licensing fees, spectrum fees, universal service obligations, or compliance requirements differ (in practice, not on paper), the result is distorted pricing and investment.
For AI-heavy e-commerce, that shows up as:
- uneven availability of affordable high-speed broadband
- slower expansion into under-served areas
- volatile costs for last-mile connectivity and business-grade internet
2) A messy playing field for AI-powered digital services
E-commerce platforms don’t operate in isolation. Your checkout, fraud checks, live chat, delivery tracking, and customer messaging depend on third parties—payment providers, logistics, cloud contact centres, and marketing tech.
If the network market shifts unpredictably, digital service providers face:
- contract renegotiations mid-term
- SLA pressure (support queues spike when connectivity is unstable)
- higher fraud exposure (attackers exploit degraded verification flows)
3) Slower time-to-value for AI projects
Batyi also highlighted a practical reality: policy processes aren’t quick. The directive itself took months to finalise, and regulatory changes can take longer.
AI adoption in retail often runs on quarterly cycles (“ship a new model, test uplift, iterate”). Regulation runs on multi-quarter cycles. If you don’t plan for that mismatch, your AI roadmap becomes hostage to external timelines.
EEIP transparency: the hidden dependency for digital trust
Answer first: Without transparent EEIP monitoring, South Africa risks replacing one barrier to entry with another—uncertainty—and uncertainty is poison for long-term network investment.
The argument about EEIPs isn’t just political. It’s about institutional capacity: who monitors commitments, how they’re audited, and what happens when targets aren’t met.
Batyi pointed to a major governance problem: oversight bodies may lack the funding and mechanisms to audit or enforce EEIP outcomes consistently.
From an AI and e-commerce viewpoint, transparency matters because trust compounds:
- Retailers invest when they can forecast network quality and price curves.
- Networks invest when regulatory enforcement is consistent.
- Consumers adopt digital services when reliability improves and prices become fair.
Here’s the stance I’m comfortable taking: South Africa doesn’t need “lighter” rules or “harder” rules as much as it needs auditable rules. If an EEIP is acceptable as a mechanism, then measuring it must be non-negotiable.
What “auditable” should look like (practically)
If I were advising a regulator or industry working group, I’d push for a simple baseline that any digital operator would recognise:
- Standardised EEIP scorecards (same categories, same definitions)
- Annual independent audits with published summaries
- Clear enforcement triggers (what happens when targets are missed)
- Public reporting that’s readable (not 200-page PDFs no one uses)
That kind of structure reduces the uncertainty premium that ultimately gets baked into connectivity pricing—pricing your customers feel.
What e-commerce leaders should do while policy plays out
Answer first: Build your AI roadmap to survive network variability and regulatory delays, then use that resilience as a competitive advantage.
You can’t control what Icasa decides, or how long amendments take. You can control how exposed your business is to those decisions.
1) Design AI features for low-bandwidth reality
If your AI personalisation makes pages heavier, you’re betting against a meaningful slice of the market.
Actions that work:
- use smaller on-device or edge-friendly models for lightweight tasks (search suggestions, text normalisation)
- compress images aggressively and treat video as opt-in
- cache recommendations and content blocks server-side
- keep checkout flows fast even when “smart” features fail
A useful internal rule: your store should degrade gracefully to “dumb but fast.”
2) Make telecom dependency visible in your metrics
Most e-commerce dashboards track conversion, AOV, CAC, and refunds. Add connectivity-aware signals:
- checkout failure rate by geography and time-of-day
- customer support contacts correlated with payment timeouts
- delivery tracking engagement vs network performance
- session abandonment after chat widget loads
If you see “network-shaped” pain, you can prioritise fixes that don’t require waiting for the market to improve.
3) Reduce single-provider risk in critical workflows
You don’t need to multi-home everything. But for the parts that directly touch revenue and trust, redundancy pays.
Consider resilience for:
- payment retries and alternate rails
- OTP and identity verification (multiple messaging routes)
- customer support channels (chat + WhatsApp + email fallback)
- fraud tooling that continues scoring even if some signals drop
4) Treat compliance and policy monitoring as a product function
This is where many teams get it wrong. They leave regulatory monitoring to legal once a quarter.
For AI-powered digital services, you want a tighter loop:
- a named owner for “policy watch” (product ops, risk, or strategy)
- monthly scenario updates (best case / base case / worst case)
- vendor contract clauses that anticipate regulatory shifts
The point isn’t to predict politics. It’s to avoid being surprised by it.
Why this matters for AI customer experience in South Africa
Answer first: The best AI customer experience is invisible—fast, reliable, and fair—and telecom policy execution determines whether “fast and reliable” is achievable at scale.
When people talk about AI in e-commerce, they focus on shiny features: chatbots, generative product copy, automated campaigns, predictive stock planning. Those are real, and we cover them across this series.
But customers judge you on basics:
- Does the page load quickly?
- Does checkout work the first time?
- Does delivery tracking update?
- Does support respond before frustration turns into churn?
AI can improve each of those. Yet every one of them depends on the same foundation: stable connectivity and a market structure that encourages investment.
The telecoms industry’s message—supportive of the directive, wary of execution—should be read as a broader warning: if the rules are unclear or unevenly enforced, the cost shows up in slower rollouts, higher prices, and weaker digital services.
If you’re building AI for e-commerce, what should you ask for?
Answer first: Ask for transparent enforcement, regulatory parity, and timelines that businesses can plan around.
Whether you’re a retailer, marketplace, or digital subscription business, your interests overlap with telecom operators more than you might think.
Three practical asks (from any industry stakeholder) are reasonable and specific:
- Transparency: publish measurable EEIP outcomes and monitoring responsibilities.
- Parity: enforce equivalent obligations where services are functionally equivalent.
- Timelines: provide realistic implementation schedules so planning isn’t guesswork.
That combination supports competition while giving investors and builders the predictability needed to keep improving networks—and keep AI features usable for real customers.
As this series keeps arguing: AI adoption in South Africa isn’t just about models and prompts. It’s about the rails—connectivity, governance, and execution—that determine who gets reliable digital services and who doesn’t.
If you’re planning your 2026 roadmap right now, the question to carry into every AI initiative is simple: are we building something that works only in perfect conditions, or something that wins in South Africa as it is?