Transnet’s IT dispute: lessons for AI-ready SA firms

R60 million is a rounding error in some boardrooms. For a loss-making state-owned enterprise, it’s a loud alarm bell—especially when that money is owed because an IT handover couldn’t happen safely.

That’s the practical headline behind the recent Transnet vs Gijima saga: Transnet must pay Gijima R60 million for services delivered between July and September 2025, even as a broader R1.5 billion dispute heads to arbitration targeted to finish by mid-February 2026. A judge’s reasoning was blunt: Transnet didn’t have a workable transition plan, wasn’t technically ready, and had no clear path to migrate the most critical component—the mainframe.

If you run an online store, a delivery app, a fintech product, or any digital service in South Africa, this isn’t “public sector drama” you can ignore. It’s a live case study in what happens when infrastructure decisions, vendor contracts, and operational readiness fall out of sync—and why AI and automation in e-commerce operations are becoming less about hype and more about survival.

What the Transnet–Gijima case really shows (beyond the headlines)

The core lesson is simple: You can’t exit a critical IT vendor relationship on paperwork alone. You need technical readiness, tested runbooks, and operational proof that you can take the keys without crashing the car.

Transnet’s contract with Gijima covered data centre operations, hosting services, enterprise IT, and mainframe services under a five-year agreement that started in December 2019 and ran to November 2024, later extended to February 2025. When Transnet tried to separate and move to a new partner, the court found a hard reality: the organisation wasn’t prepared to accept handover.

For private businesses, the same dynamic shows up in different clothes:

• An e-commerce retailer wants to switch cloud providers before peak season but hasn’t validated payment flows and integrations.
• A marketplace changes its ERP or warehouse management partner, then discovers stock accuracy and order routing were “held together” by undocumented scripts.
• A digital services provider plans a CRM migration without a complete data map, and then customer support collapses for a week.

The result is always the same: delay, cost overruns, and the dreaded “keep paying while we figure it out.”

The mainframe point matters more than it sounds

Many modern businesses don’t run mainframes—but the pattern is identical. In Transnet’s case, the mainframe was described as the most critical component and the hardest to migrate.

In e-commerce and digital services, your “mainframe equivalent” is usually one of these:

• Payments and reconciliation
• Customer identity and authentication
• Order management (OMS) logic
• Product and pricing “source of truth”
• Warehouse and courier integrations

These systems are rarely replaced cleanly. They’re replaced in pieces, with careful parallel runs and constant monitoring. If you don’t know what your “cannot fail” component is, you’re already behind.

Why contract disputes are often a symptom of operational fog

Contract conflict tends to start long before lawyers get involved. It starts when the business can’t answer basic operational questions quickly and confidently.

Examples I see repeatedly:

• “What exactly are we running?” Inventory of apps, services, APIs, batch jobs, and dependencies is incomplete.
• “Who owns what?” Responsibilities between internal teams and vendors are blurry.
• “Are we ready to transition?” There’s no agreed definition of “ready,” no exit checklist, and no proof via testing.
• “What are we paying for right now?” Costs are buried in invoices rather than tied to measurable service outcomes.

In the Transnet case, the court commentary highlights the absence of a transition plan and technical readiness. That’s not a legal nuance. It’s operational hygiene.

Arbitration as a business cost, not just a legal process

Arbitration is often framed as “we’ll settle it later.” The operational reality is harsher: while a dispute is unresolved, teams hesitate, projects slow down, and vendors minimise risk by sticking strictly to contract terms.

For e-commerce companies, that stall is expensive during South Africa’s busiest trading windows—Black Friday hangovers, festive season returns, and January back-to-school demand. Downtime, delayed features, or shipping issues aren’t theoretical; they show up as:

• Higher cart abandonment
• More refunds and chargebacks
• Support backlogs
• Courier penalties
• Lower repeat purchase rates

Where AI actually helps: reducing the “unknowns” that derail transitions

AI won’t magically prevent conflict, but it does reduce the uncertainty that makes transitions risky and disputes more likely.

Here are the most practical ways AI in e-commerce and digital operations can keep you out of the Transnet-style trap.

1) AI-assisted application discovery and dependency mapping

Answer first: You can’t transition what you can’t see.

AI-powered discovery tools can ingest logs, configuration data, network flows, and code repositories to map:

• Which services call which APIs
• Where customer and order data travels
• Which batch jobs update stock, pricing, or invoicing
• Which integrations are most failure-prone

This is the foundation for a realistic transition plan—especially when institutional knowledge is thin or staff turnover is high.

2) Automated runbooks and “day-2 operations” hygiene

Answer first: Most migrations fail after go-live, not during it.

Automation platforms can turn tribal knowledge into repeatable runbooks:

• Standard incident responses (cache clear, queue replay, failover steps)
• Scheduled checks (payment gateway health, courier API latency)
• Safe deployment steps with rollbacks

Add AI copilots that summarize incidents and propose next actions, and you reduce mean time to resolution when things wobble—crucial during any vendor transition.

3) Contract intelligence for IT and vendor management

Answer first: Many disputes come from mismatched expectations hidden inside PDFs.

AI can help vendor managers and procurement teams extract and compare:

• Service level agreements (SLAs) vs actual performance
• Exit clauses and transition assistance obligations
• Definitions of “handover,” “disengagement,” and “acceptance”
• Billing triggers and invoice validation rules

This matters for digital services in South Africa, where vendors often provide blended services across cloud, hosting, development, and support. If your contract language is vague, your operational plan must be extra specific—or you’re paying for ambiguity later.

4) Predictive monitoring for customer-impacting systems

Answer first: The best time to detect failure risk is before customers notice.

AI-based observability can detect early signals such as:

• Rising checkout latency after a deployment
• Increased payment retries by bank or card type
• Courier label failures by route or warehouse
• Abnormal spikes in refunds for a single product category

When you’re migrating platforms or replacing vendors, these signals tell you whether your “new world” is behaving like the old one—without waiting for angry customers.

A practical playbook for AI-ready transitions (what I’d do before switching vendors)

Most companies get this wrong by treating transition as an IT project, not a business continuity program.

Here’s a straightforward approach that fits e-commerce, marketplaces, and digital service providers.

Step 1: Define “technical readiness” in measurable terms

Write down acceptance criteria that a non-technical exec can still understand. For example:

• 99.9% successful payment authorisations in parallel run for 14 days
• Inventory accuracy within ±1.5% for top 500 SKUs
• Order-to-dispatch time not worse than baseline by more than 10% during peak hours
• Support tickets per 1,000 orders not exceeding baseline

If you can’t measure readiness, you can’t defend it in a board meeting—or in a dispute.

Step 2: Identify your “mainframe equivalent” and isolate it

Pick the system you cannot afford to break. Then:

• Reduce custom logic around it
• Document data contracts (inputs/outputs)
• Build a parallel run strategy
• Add extra monitoring and rollback plans

This is where AI-powered dependency mapping and observability earn their keep.

Step 3: Build an exit plan before you need it

A vendor exit plan should exist on day one, not at contract end. Include:

• Data export formats and timelines
• Credential and key rotation steps
• IP, configuration, and documentation handover
• Named responsibilities (RACI) across both parties
• A tested cutover plan with rehearsal dates

If you wait until the relationship is tense, cooperation drops fast.

Step 4: Use AI to create a single source of operational truth

Create an “ops brain” that combines:

• System inventory and ownership
• Monitoring dashboards and alert history
• Incident summaries and postmortems
• Vendor performance vs SLAs
• Cost-to-serve metrics (per order, per ticket, per shipment)

This directly supports better decision-making and reduces the “he said, she said” dynamic that fuels arbitration.

People also ask: can AI prevent IT contract disputes?

AI can’t prevent disputes on its own, but it can reduce the main causes: unclear scope, weak evidence, and poor readiness.

A dispute usually turns on questions like “who was responsible,” “what was delivered,” and “was the client ready to accept handover.” AI helps you keep cleaner records (tickets, changes, logs), clearer performance evidence (SLAs vs actuals), and more realistic transition plans (dependency maps, parallel run monitoring).

If you’re serious about using AI for e-commerce operations in South Africa, start here: use AI to make your operations legible. Clarity beats confidence.

What this means for South Africa’s digital economy—and your roadmap for 2026

Transnet sits underneath a big chunk of South Africa’s logistics reality. When core infrastructure organisations struggle with IT transitions, the ripple effect touches the broader digital economy: supply chains, delivery reliability, and the cost base that ultimately hits retailers and consumers.

For private companies, the takeaway is less about Transnet’s specifics and more about the recurring pattern: complex IT environments punish vague transition planning. The businesses that win in 2026 won’t be the ones with the flashiest AI demos. They’ll be the ones using automation and AI to keep platforms stable, costs controlled, and vendor relationships clean.

If you’re planning a platform migration, a new cloud partner, or a major systems upgrade, start by stress-testing your transition readiness—then use AI to close the gaps. What would break first in your business if your current IT partner handed you the keys tomorrow morning?