African Utilities Are Trying to Run AI on Infrastructure Built for Spreadsheets. Here Is How to Fix It.

Every time I begin an analytics engagement with a new energy organisation, I do the same assessment first. Not of their analytics capability — of their data infrastructure. In fifteen years, I have not yet encountered an African utility whose data infrastructure was adequate to the AI ambitions of its leadership. The gap between what organisations want to do with AI and the data infrastructure they have to support it is, in most cases, the primary constraint on their progress.

This is not a technology problem. It is an investment prioritisation problem. Data infrastructure is invisible — it does not produce outputs that can be presented to a board, it does not generate excitement at conferences, and its value is indirect and realised over years rather than immediately. These characteristics make it systematically underfunded relative to the AI applications that depend on it.

The Four Layers of Data Infrastructure That African Utilities Need

Layer 1: Data Collection Infrastructure

The foundation is instrumentation — the sensors, smart meters, SCADA systems, and IoT devices that convert physical operational reality into digital data. Most African utilities have some of this infrastructure but not enough of it, and what they have is often poorly maintained, poorly calibrated, and generating data of questionable quality. The priority investment here is not the newest technology — it is ensuring that existing instrumentation is working correctly and that the data it generates is accurate and complete.

Layer 2: Data Integration Infrastructure

African utility organisations typically have operational data spread across multiple systems — a billing system, a customer management system, a work order management system, a geographic information system, a SCADA platform, and frequently a collection of departmental spreadsheets that operate entirely outside any formal system. These sources do not automatically talk to each other. Building the integration infrastructure — the data pipelines, common identifiers, and reconciliation processes that allow data from different sources to be combined — is the prerequisite for any cross-system analytics.

The organisation that cannot answer the question "which customers are affected by this fault right now?" — because customer location data is in one system, network topology is in another, and there is no integration between them — is not in a position to benefit from AI. The AI has nothing to work with. The integration infrastructure is the work that needs to happen first.

Layer 3: Data Storage and Processing Infrastructure

Modern analytics and AI require the ability to store large volumes of operational data over time and process it efficiently. Most African utility organisations are storing their operational data in systems designed for transaction processing, not analytical processing. Separating analytical workloads from operational workloads — through a data warehouse, a data lake, or a cloud analytics platform — is necessary to support the volume and complexity of analytics required for AI deployment.

Layer 4: Data Governance Infrastructure

The least visible and most underinvested layer is governance: the processes, standards, and accountability structures that ensure data quality is maintained over time. Without governance infrastructure, the improvements produced by data quality initiatives degrade as operational teams return to old practices. Governance infrastructure includes data ownership assignments, data quality standards, audit processes, and the performance management integration required to make data quality a management priority rather than an IT afterthought.

How to Make the Investment Case

The business case for data infrastructure investment is most compelling when expressed as the constraint on value from AI programmes that have already been approved or are being considered. If the organisation has committed to an AI-powered revenue assurance programme that requires clean, integrated metering data — and the data is not clean or integrated — then the cost of the data infrastructure investment is the cost of realising the value already committed to. Framing it this way transforms data infrastructure from a cost centre request to a prerequisite for value already promised.

Dr. Sunny Okonkwo

AI Strategist · Head of Data Analytics at one of Africa's largest energy and utility companies. Author of 7 books including the #1 Bestseller The AI Alchemist. Keynote speaker at IIBA, Big Data Summit Canada, Global Summit, and UNICAF.