The Sequence Problem Nobody Is Talking About

A quiet pattern is emerging across South African cold chain discourse in early 2026. Researchers are documenting the waste our industry generates. Gartner is forecasting how AI will reshape supply chain workforces. Technology commentators are describing a future where IoT-connected goods flow seamlessly across African borders, guided by machine learning and real-time data.

The analysis is sharp. The data is credible. And if you’re not careful, you’ll walk away believing the solution is to implement AI.

It isn’t.

The problem — the one that connects all three threads — is that South Africa’s cold chain infrastructure is not yet capable of generating the consistent, reliable data that AI requires to function. We are proposing to build intelligence on top of a foundation that is still fragmented, underpowered, and cold-blind at critical points in the chain. That is not a technology adoption challenge. It is a sequencing problem.

Get the sequence wrong, and AI becomes expensive decoration on a broken system. And as further ColdChainSA analysis makes clear, the sequencing error carries two additional costs that the global AI narrative ignores entirely: the system can be attacked at the moment it is most dependent on digital infrastructure, and there aren’t enough trained people in South Africa to run it even if the infrastructure were ready.

This article argues for getting the sequence right — and for understanding what “right” actually requires in this market.

What the Research Actually Shows

18,000 Tonnes of Produce Lost — Not at the Farm, Not at the Shelf

Recent doctoral research from Stellenbosch University, conducted by food systems researcher Dr. Ikechukwu Opara, has identified a specific and significant waste problem at South Africa’s wholesale fresh produce markets. The research found that between 9,124 and 17,969 tonnes of fruits and vegetables are wasted annually at these wholesale hubs — not on farms, not in retail — at the wholesale market level, where produce moves between primary distribution and retail supply chains.

To contextualise that figure: the upper estimate represents approximately 900 fully loaded large trucks of food being discarded every year, in a country where more than 63% of households experience varying levels of food insecurity.

Dr. Opara’s research identifies the most critical failure point not as storage, but as the receipt of produce. Operational delays routinely leave crates of fresh fruit sitting at ambient temperature for hours before transfer to cold storage. During South African summer, when markets operate at peak volume and ambient temperatures are highest, the combination is particularly destructive. Transportation in unrefrigerated vehicles compounds the damage, subjecting produce to unfavourable humidity conditions that accelerate respiration and shrivelling — resulting in commercial loss that flows back to producers and forward to consumers as price pressure.

Dr. Opara’s proposed solution includes AI-enabled monitoring: automated systems that alert staff when temperature or humidity deviates from acceptable ranges, demand forecasting tools that optimise ordering to prevent oversupply, and dynamic pricing algorithms that track remaining shelf life to accelerate sales of produce approaching quality limits.

These are legitimate applications. The question is not whether they work — in instrumented, connected environments, they demonstrably do. The question is what must exist before they can work. The wholesale market interface is the least instrumented point in the chain. It is also where the temperature breaks are actually occurring. You cannot apply AI-driven monitoring to a space that has no monitoring to begin with.

Gartner’s Workforce Warning: High Performers Reinvent, Others Just Cut

A February 2026 Gartner survey of 509 supply chain leaders globally found that more than half (55%) expect advancements in agentic AI to reduce the need to hire for entry-level positions, with 51% believing the technology will drive overall workforce reductions.

What is less widely cited is the finding that most meaningfully distinguishes high-performing organisations from the rest. High performers — those that exceeded expectations across customer lead time, satisfaction, time to market, revenue growth and sustainability metrics — are not treating AI as a headcount reduction tool. They are using it to redesign how work gets done. Their talent strategies focus on upskilling for the AI era and increasing automation to improve efficiency, not simply eliminating positions.

Gartner’s Marco Sandrone is direct: “The highest performing supply chain organisations are using AI to reinvent how work gets done and how talent is developed. They are not treating AI as a blunt instrument for headcount reduction.”

For South African cold chain operators, this distinction matters for a reason the global survey cannot capture. As ColdChainSA’s workforce analysis documented in Cold Chain Workforce Crisis: South Africa’s Invisible Skills Gap, the country already faces a structural shortfall: the cold chain market is projected to grow from $6.3 billion to $20.6 billion by 2030, yet South Africa produces roughly 15,000 artisans annually against an economy-wide need for approximately 30,000. Cold chain workers don’t even appear as their own category in official workforce data — they are buried under “transport,” “agriculture,” or “manufacturing,” invisible to the planning frameworks that should be developing them.

Against this backdrop, the Gartner finding takes on a different weight. The temptation for mid-sized cold chain operators under margin pressure will be to implement AI as a cost-reduction mechanism — displacing entry-level monitoring and coordination roles before the operational infrastructure is ready to support AI-driven decision-making. This is precisely the pattern Gartner associates with underperformers. And in the South African context, it is not just strategically misguided: it eliminates the manual capability that is currently the only backup when digital systems fail.

IoT and the Invisible Bottleneck Africa Refuses to Own

The third piece of the picture comes from a February 2026 ITWeb opinion piece by Peter Walsh, MD of CommsCloud. Walsh’s argument is structural: Africa’s trade bottlenecks are routinely described as physical infrastructure problems because these are visible and politically tangible. The actual damage sits beneath the surface in digital connectivity architecture that was never designed to work across borders at scale.

For cold chain operators, the implications are direct. A single cross-border truck generates tens of thousands of data points per hour — location, temperature, video feeds, engine diagnostics, driver behaviour. When that data flows seamlessly across borders, trade fundamentally changes. But this requires resilient, uninterrupted connectivity — and that is precisely what current IoT infrastructure in African logistics cannot consistently deliver.

Walsh identifies a structural cause: for years, the industry has relied on roaming SIMs and centralised networks as necessary compromises. A 2025 Kaleido Intelligence survey found that only 2% of IoT operators still believe global roaming SIMs work for large-scale deployments. Governments are tightening controls on permanent roaming. The old infrastructure model is ending — not because the industry evolved gracefully, but because it has run out of excuses.

Walsh’s core point deserves to be stated plainly for a cold chain audience: every time cold chain data vanishes between countries, every time monitoring systems go dark at borders, the impact lands on operators — not on the systems that failed them. This is the invisible bottleneck. And it is not unique to cross-border operations. The same connectivity fragility affects domestic operations — refrigerated vehicles that lose monitoring visibility in coverage-weak corridors, cold storage facilities that experience data gaps during load shedding, temperature loggers that fail to sync when network infrastructure is unreliable.

The Two Compounding Risks South Africa Adds

When the Algorithm Fails: Cybersecurity Is the AI Risk Nobody Is Pricing In

The global AI narrative for cold chain focuses almost entirely on what happens when systems work. ColdChainSA’s analysis of AI dependency and human backup plans focuses on what happens when they don’t.

AI isn’t magic. It’s software — running on servers that need electricity, connected through networks that can be interrupted, processing data that can be corrupted or falsified. Every AI system in the cold chain has the same fundamental vulnerability categories as any other software: it can be hacked, fed false data, make wrong decisions from correct data, or simply go offline. The difference in cold chain is that any of these failure modes can result in product loss, compliance violations, or harm to end consumers — all on a time-critical basis.

The attack surface is not theoretical. In March 2025, Astral Foods — South Africa’s largest chicken producer — suffered a cyberattack that disrupted operations for an entire week, costing more than $1 million in losses. The Transnet ransomware attack in July 2021 paralysed the ports of Durban and Cape Town severely enough that Transnet declared force majeure. Shoprite Group faced a RansomHouse attack in 2022 that compromised customer data across Africa’s largest retail operation.

Internationally, ransomware attacks on food and agriculture surged 100% in early 2025, with 84 incidents in Q1 alone. The Blue Yonder supply chain software attack in November 2024 disrupted Starbucks’ scheduling across 11,000 stores and affected UK grocery operations. A single attack on Cencora compromised 27 pharmaceutical and biotechnology companies simultaneously.

The pattern is consistent: as AI takes on more decisions in cold chain operations, human backup knowledge disappears. Operators who once knew how to run manual processes retire. New staff are trained on digital systems exclusively. When those systems fail — not if — there is nobody left who knows how to keep product moving.

South Africa has a narrow window here worth naming explicitly. Many local operations are still in early digitisation stages. The informal processes and manual capabilities still exist. Load shedding has, perversely, provided years of practical resilience training: every South African cold chain operator has lived through unplanned system resets, unpredictable outages, and forced fallback to manual processes. UPS systems that work when tested, generator fuel maintained, manual temperature monitoring procedures, communication that functions without internet — these load-shedding protocols are essentially a rehearsal for a cyberattack response. Operators who have invested in this infrastructure are better prepared than they may realise.

The question is whether this manual capability will be formalised and preserved as AI adoption accelerates — or whether it will be systematically eliminated in the name of efficiency before the digital systems are mature enough to operate safely without it.

The Workforce Problem Makes Everything Else Worse

The sequencing argument — infrastructure before AI — assumes there are people available to build and operate the infrastructure. In South Africa’s cold chain, that assumption requires examination.

ColdChainSA’s Digital Skills Gap analysis maps a 7-layer cold chain technology stack: physical operations; cold chain operations protocols; IoT devices and sensors; connectivity architecture; platform and cloud infrastructure; analytics and AI; and compliance systems. Traditional artisan training addresses the first two layers at best. The supply chain transformation that Gartner describes operates primarily at layers five and six. The gap — layers three and four, the bridge between physical operations and digital intelligence — is where the workforce deficit is most acute and least visible. More than 60% of IoT-related roles are already unfilled in South Africa, and 73% of organisations expected significant skills gaps by 2025. Cold chain operators compete for this same limited talent pool while requiring additional domain expertise that generic technology workers simply don’t have.

The industry’s 69% cybersecurity incident rate compounds this directly. When an attack hits a cold chain operation, you need cybersecurity expertise to contain the breach, cold chain expertise to assess product safety during the outage, compliance expertise to document what happened, and operational expertise to keep things moving on manual procedures. Finding one of these skill sets in South Africa’s constrained talent market is challenging. Finding all four in coordination during a crisis is close to impossible.

The Gartner high-performer finding — that leading organisations are upskilling for the AI era rather than reducing headcount — requires a workforce capable of upskilling. South Africa’s cold chain does not yet have that workforce in sufficient numbers. The AI tools are arriving into an industry that cannot find enough qualified people to operate its current equipment, let alone the new category of AI-literate, data-fluent cold chain technicians the technology era demands.

This is not an argument against AI adoption. It is an argument that workforce investment must happen in parallel with infrastructure investment, not sequentially after it — and that both must happen before AI tools are deployed as operational decision-makers rather than decision-support tools.

The South African Context Makes This More Acute

South Africa’s cold chain faces compounding challenges that generic global analysis does not capture.

• Cold chain breaks are irreversible, not just inconvenient. When the temperature chain is interrupted — whether at a wholesale market loading dock or on a reefer container waiting at port — straightforward physics takes over. For every 10°C rise in temperature, the rate of biological activity in fresh produce roughly doubles (the Q10 effect). A pallet of table grapes moved from 0°C to 20°C respires at four times the intended rate, consuming its own energy reserves, producing heat that raises ambient temperature further, and releasing ethylene that accelerates ripening in surrounding product. This cascade cannot be reversed by restoring cold temperatures. The remaining shelf life is permanently shortened. As ColdChainSA’s analysis of the financial cost of cold chain failure documents, this physics translates directly to measurable losses: R350 million in direct industry losses and R1 billion in at-risk inventory in a single export season at Cape Town port alone. AI systems that predict spoilage cannot operate on data that was never collected during the breaks where the damage actually occurred.
• Load shedding degrades the cold chain at both ends. Cold storage facilities that experience power interruptions face thermal cycling that affects produce quality and creates temperature excursions that may or may not be accurately recorded depending on monitoring system battery life. The data that does get captured is often incomplete, inconsistent, or flagged with anomalies that corrupt AI training datasets.
• Altitude affects equipment performance in ways that are poorly documented. Gauteng operates at approximately 1,700 metres above sea level. Refrigeration equipment at altitude delivers 15–20% less cooling capacity than rated specifications. When AI-driven predictive maintenance is applied to this equipment, it will inherit baseline data that reflects altitude-derated performance as normal — unless the underlying data collection is designed to capture this variable explicitly.
• The wholesale market gap is a data desert. Dr. Opara’s research identifies the wholesale market as the critical failure point. From a data perspective, it is also where the least instrumentation currently exists. This is where temperature breaks are occurring, and it is the gap in the data record that makes AI-driven quality prediction unreliable downstream.
• The SME sector operates largely without monitoring infrastructure. While major operators like Vector Logistics and CCH operate sophisticated monitored fleets, a significant portion of South African cold chain capacity — particularly in last-mile delivery — operates without continuous temperature monitoring, automated alerting, or digital record-keeping. Any AI implementation targeting industry-wide waste reduction must contend with this fragmentation.

What Infrastructure First Actually Means in Practice

Arguing that infrastructure must precede AI is not an argument against technology adoption. It is an argument about sequencing — and about avoiding the waste that comes from implementing sophisticated tools on inadequate foundations.

• Continuous temperature monitoring is non-negotiable before AI monitoring can exist. For operators without SANAS-calibrated temperature loggers generating uninterrupted records across vehicles and facilities, the first technology investment is not AI — it is basic monitoring infrastructure. The monitoring creates the data; the AI analyses it.
• Connectivity architecture must be designed for reliability, not convenience. If temperature logger data is transmitted via a single-carrier roaming SIM that loses signal in corridors your vehicles regularly travel, you are generating intermittent data records with gaps at precisely the points where the cold chain is most vulnerable. Moving to multi-IMSI or eSIM-based connectivity that maintains signal continuity is an infrastructure investment that precedes any meaningful AI application.
• Manual capability must be preserved alongside digital systems, not displaced by them. The P&G response to the Blue Yonder attack — manual backup systems operational within 12 hours — represents the resilience standard the industry should target. This requires deliberately maintaining paper-based procedures, training staff on manual protocols quarterly, and conducting regular digital blackout drills. South Africa’s load shedding experience already provides the rehearsal infrastructure for this. The failure would be to let that capability atrophy as AI systems assume more operational responsibility.
• Workforce capability must be built alongside technology. Cold chain operations that rush to automate before developing internal capability to interpret and act on AI-generated insights will find that the technology underdelivers. Building data literacy and AI-oversight capability in parallel with technology implementation — not as a consequence of it — is what distinguishes organisations that succeed with AI from those that merely acquire it.

The 2035 Window Is Real — But Requires Deliberate Action Now

Walsh frames the challenge as a continental choice: Africa can keep layering trade agreements, logistics platforms and AI tools on top of fragile, fragmented digital infrastructure, or it can redesign the foundation. The 2035 horizon he describes is not abstract for South Africa’s cold chain industry.

The R53 billion fresh produce market, the expanding pharmaceutical cold chain sector, and the growing cross-border trade corridors to SADC markets all represent genuine opportunities. The question is whether the industry will be positioned to capture them with AI-enabled operational intelligence — or whether it will still be losing 18,000 tonnes of produce annually at wholesale markets that have not yet been connected to the monitoring infrastructure that makes AI possible, while simultaneously vulnerable to cyberattacks that can take the entire system offline, and unable to find enough qualified people to operate the equipment it already has.

The AI is coming. But it will arrive to find either a cold chain that has built the infrastructure to use it safely — monitored, connected, cyber-resilient, staffed by people trained across the full skills stack — or one that is still generating the incomplete, inconsistent data record that has characterised operations for the past decade, one ransomware attack away from paralysis, with no manual fallback and no trained workforce to respond.

The sequence matters. Infrastructure first. Resilience built in. Workforce developed in parallel. AI second.

Sources & References

Research & Academic Sources

• The Cold Reality: AI and Infrastructure Reform Could Save 18,000 Tonnes of SA Produce — Agri News, February 25, 2026. Coverage of Dr. Ikechukwu Opara’s Stellenbosch University doctoral research on postharvest losses at South African wholesale fresh produce markets.

Industry Research & Analysis

• Agentic AI Expected to Reduce Entry-Level Supply Chain Hiring — IT-Online, February 26, 2026. Gartner survey of 509 global supply chain leaders on agentic AI workforce impact and high-performer talent strategies.
• Trade Doesn’t Flow on Roads — It Flows on Data — ITWeb, February 26, 2026. Peter Walsh (CommsCloud) on IoT connectivity infrastructure failures in African trade logistics and the roaming SIM architecture problem.
• Kaleido Intelligence Report: Regulatory Compliance Is the Number One Connectivity Challenge for IoT Providers — floLIVE / Kaleido Intelligence, 2025. Only 2% of IoT operators consider global roaming SIMs viable for large-scale deployments.

ColdChainSA Analysis

• When the Algorithm Fails: Why AI-Dependent Cold Chains Need Human Backup Plans — ColdChainSA, February 24, 2026. Cybersecurity attack surface analysis, SA-specific vulnerabilities (Astral Foods, Transnet, Shoprite), resilience framework, and the load shedding–cybersecurity analogy.
• Cold Chain Workforce Crisis: South Africa’s Invisible Skills Gap — ColdChainSA, February 20, 2026. Seven converging demand forces on SA cold chain workforce capacity, the 7-layer skills stack, and the data black hole that makes cold chain workers statistically invisible.
• The Digital Skills Gap Nobody’s Measuring: What Cold Chain Operations Will Need by 2030 — ColdChainSA, March 2026. Detailed mapping of the 7-layer cold chain technology stack, role-by-role skills requirements through to 2030, and the gap between skills currently being trained and skills the industry will actually need.
• The Cost of Cold Chain Failure: How Broken Links Are Costing South Africa Billions — ColdChainSA, February 2026. Quantified analysis of SA cold chain failure costs including the CTCT port crisis, Q10 thermodynamic effects, and the R21.7 billion annual household food waste figure.
• South Africa’s Cold Chain Infrastructure Transformation — ColdChainSA, 2025. Background on major cold chain investment programmes and infrastructure development across South African logistics corridors.

Directory Resources

• Cold Chain Technology & Monitoring Directory — ColdChainSA. Verified temperature monitoring, IoT and cold chain technology suppliers operating in South Africa.
• South Africa Refrigerated Transport Directory — ColdChainSA. Verified refrigerated transport operators across South African provinces.