By: Junaid Kleinschmidt – Intelligence lead and Ntsako Baloyi – Data & AI lead for Accenture, South Africa

Within the South African business community, cloud adoption has moved well beyond early migration conversations. Yet many organisations still treat cloud as a completed journey once systems are stable and uptime targets are met. The reality is more complex. Cloud is no longer a destination, but a continuously evolving foundation that must support data, AI and business change in real time. As AI investment accelerates globally, the ability of local organisations to extract value from these technologies will depend less on ambition and more on the strength of their underlying cloud and data environments.

At the same time, it is important to be clear that organisations do not need cloud to do AI. AI workloads can run on-premise or in hybrid environments. However, cloud has become the dominant narrative because it enables faster scaling, easier access to tools and reduced upfront infrastructure investment. The risk is when this narrative is accepted without considering whether it aligns to a company’s specific operating model, regulatory environment or cost profile.

Recent insights show that while most organisations have made progress in migrating workloads, only a small proportion are truly ready to support AI at scale. Many remain constrained by legacy systems, fragmented data environments and inconsistent governance practices. Telecommunications have generally advanced further, driven by compliance and competition, followed by Resources, Retail, Consumer Goods and Automotive, while other industries are still formalising their cloud strategies and operating models.

One of the more persistent challenges is the misconception that cloud adoption automatically translates into innovation. In practice, organisations that focus only on migration often find themselves dealing with rising costs, limited visibility and growing technical complexity. Without clear alignment between cloud investments and business outcomes, cloud can become another operational burden rather than a driver of growth. This is particularly relevant in a local context where budgets are scrutinised closely and technology spend must demonstrate tangible value.

A growing concern is the risk of “cloud bill shock”, particularly as AI usage increases. Large language models are typically priced on a consumption basis, with costs driven by the number of tokens processed (both input and output), meaning expenses scale with usage

Large language models are usually billed based on usage, with costs determined by the number of tokens processed – both input and output. Expenses increase as usage grows. Meaning, the more queries, prompts and integrations an organisation runs, the higher the cost. Without proper controls, this can escalate quickly. Organisations need to put guardrails in place early – including usage monitoring, cost allocation, token-optimisation, model selection strategies and clear policies on where and how AI is used. Cloud FinOps practices become critical here, ensuring that AI consumption is visible, managed and tied to business value rather than growing unchecked.

At the same time, AI is reshaping expectations. It is no longer confined to isolated use cases or pilot projects. Increasingly, it is being embedded into workflows, decision-making and customer experiences. This shift places new demands on cloud environments. Data must be accessible, governed and connected across the organisation. Systems must be able to process information in real time rather than in batches. Security and identity controls must be consistent across hybrid environments. Without these foundations, AI initiatives risk becoming fragmented and difficult to scale.

Cloud providers are playing a significant role in accelerating AI innovation, having made substantial investments in infrastructure, tooling and model ecosystems. At the same time, open-source alternatives are emerging rapidly, offering organisations the ability to experiment and deploy AI capabilities locally at a lower cost. While this creates opportunity, it also introduces risk. Not all open-source models meet enterprise-grade security, governance or compliance requirements. Organisations need to strike a careful balance between accessing innovation and maintaining control, particularly in regulated environments.

Data remains a critical constraint. Many organisations continue to operate with siloed datasets, limited integration and incomplete governance. While structured data is often prioritised, unstructured data, which is essential for advanced AI use cases, is frequently underutilised. In South Africa, where organisations often operate across multiple systems and environments, the challenge is not only about migrating data to the cloud, but ensuring it is usable, trusted and aligned to business needs.

Cybersecurity is another area where gaps are becoming more visible. As AI adoption increases, so too does the attack surface. Yet only a small percentage of organisations have fully integrated, real-time security across their cloud and on-premise environments. This creates exposure at a time when threats are becoming more sophisticated and automated. For local businesses, this reinforces the importance of embedding security into cloud design from the outset, rather than treating it as an add-on.

Cost challenges are not limited to AI consumption. Many large organisations that have migrated workloads to the cloud at scale have not applied strong Cloud FinOps practices or fully embraced cloud-native modernisation. As a result, they carry legacy inefficiencies into the cloud, making it more expensive than anticipated. In some cases, this has led to workloads being repatriated back to on-premise environments where costs are perceived to be more controllable. The balance lies in being deliberate – modernising where it creates value, optimising continuously and avoiding a one-size-fits-all migration approach.

A useful way to understand progress is to look at three broad maturity levels. Some organisations are still stabilising their environments, dealing with legacy constraints and limited automation. Others have optimised certain areas, achieving efficiency but not yet unlocking new growth. A smaller group is beginning to innovate at scale, using cloud as a platform for continuous reinvention. In South Africa, all three stages are visible, often within the same industry, reflecting different starting points and strategic priorities.

Moving forward requires a shift in mindset. Cloud decisions should be anchored in business value, not just technical outcomes. This means linking cloud investments to measurable objectives such as cost control, resilience, revenue growth or customer experience. It also requires stronger governance, clearer accountability and better visibility into how resources are used. Practices such as Cloud FinOps are becoming increasingly relevant, helping organisations manage cloud spend in a way that aligns with business performance.

Equally important is the need to modernise across the entire technology landscape. For many organisations, this will involve a mix of on-premise, private and public cloud environments. Rather than forcing a single approach, the focus should be on building capabilities that work across this continuum. This includes adopting cloud-native tools, improving observability and enabling more agile ways of working. In the South African context, this flexibility is essential, given the varying regulatory, operational and infrastructure requirements across sectors.

The pace of change across the AI ecosystem also reinforces the need for flexibility. Model providers are continuously improving performance, cost efficiency and capabilities, but no single model is best suited to every use case. At the same time, enterprise software providers are embedding AI directly into their platforms, changing how organisations consume technology. This creates a need for openness and interoperability, allowing businesses to adopt a best-of-breed approach that aligns with their strategy, skills and maturity.

There is also a growing recognition that talent and operating models need to evolve alongside technology. AI and cloud are not purely technical domains. They require collaboration between business, engineering, data, legal, and risk teams. Building consistent, cross-functional teams and investing in skills development will be key to sustaining progress. Without this, organisations risk relying on short-term interventions that do not translate into long-term capability.

Ultimately, the pace of change leaves little room for hesitation. Global data shows that a significant majority of executives plan to increase AI investment in the near term, with many viewing it as a driver of revenue growth rather than just cost efficiency. For South African organisations, the question is not whether to invest, but how to ensure that these investments deliver meaningful outcomes.

Cloud remains one of the most effective enablers of this objective when applied with discipline. Strengthening the foundation, improving visibility and embedding AI into the flow of work will allow organisations to move with greater confidence. From there, each step forward becomes easier to sustain and build upon.