Sivuyise Ndzendze, Sasol.
Traditional data warehouses are no longer sufficient for the real-time demands of modern heavy industries like mining and fuel production, said Sivuyise Ndzendze, data engineering lead at Sasol.
Speaking at the ITWeb Data Insights Summit 2026, Ndzendze explained that organisations must transition from “warehouse thinking”, which focuses on historical reporting and structured data, to “platform thinking” that enables real-time insights and handles unstructured data.
In environments like Sasol’s mines, where data must be processed in real-time, latency can have life-or-death consequences: “When a miner is in trouble, you don’t want to wait for tomorrow, you want to know now,” he stated, adding that a modern platform must provide “real-time awareness” to allow for immediate safety actions and machine shutdowns.
Ndzendze noted that Sasol deals with a massive volume of data sources, requiring a shift towards cognitive approaches rather than simple descriptive reporting.
To manage this complexity, Ndzendze outlined the roles of different architectural approaches. He described the role of lakehouse architecture as the foundational platform used to scale analytics and AI, while handling a variety of data types, such as images from lab systems used to check coal quality.
He then introduced Microsoft Fabric as an end-to-end software as a service solution that simplifies the user experience. According to Ndzendze, Fabric allows business users to see insights faster without “platform scrolls” or high operational overheads, embedding intelligence tools directly into the workflow.
When it comes to organisational strategy, Ndzendze said data mesh is a suitable operating model for scaling data ownership across larger enterprises. This approach shifts accountability closer to the business domains such as mining, procurement or finance, rather than keeping it centralised in a single IT department.
He warned that without distributed ownership and proper governance, organisations end up with “ungoverned AI” and “shadow models”, where employees revert to using Excel spreadsheets because they do not trust the central systems.
Ndzendze explained that lakehouse architecture, Microsoft Fabric and data mesh architecture are not competing concepts, but are complementary. He said a future-ready strategy uses lakehouse as the foundation, Fabric as the provider of intelligence, and data mesh as the operating model to drive the entire system.
“The future of data and AI will be decided by deliberate choices,” he concluded, emphasising the need for enforced governance standards to ensure data remains a trusted institutional asset.
