South African enterprises are rapidly integrating generative AI into their operations, but most are doing so without formal strategies, dedicated leadership or the infrastructure required to maximise value and minimise risk.

This is the key finding of the newly released South African Generative AI Roadmap 2025, based on a study by World Wide Worx in collaboration with Dell Technologies and Intel.

The report, which surveys over 100 midsized and large enterprises across industry sectors, shows that gen AI adoption has climbed from 45% of large enterprises in 2024 to 67% in 2025.

This dramatic rise positions gen AI as the fastest-moving digital trend in the country. However, in a rush to adopt the fast-growing technology, there is a need for organisations to take the foundational steps of planning and governance. Doing so will more clearly connect AI to people and processes and help organisations reap genuine, sustaining ROI.

“Many organisations are simply unaware of the gaps they’re leaving in their systems,” says Arthur Goldstuck, CEO of World Wide Worx and principal analyst of the study. “The risk goes beyond the technical and includes reputational, ethical and operational vulnerability. While the first step of technology adoption is well under way, our survey demonstrates there is room for operational growth.”

According to the report’s findings, AI adoption has brought clear benefits to the organisations using it:

Eighty-six percent of gen AI users cite increased competitiveness as a result of using AI tools;
Eighty-three percent report improved productivity; and
Sixty-six percent see enhanced customer service.

Yet, behind these numbers lies an operational gap:

Only 14% of organisations have a formal company-wide gen AI strategy;
Just 13% have implemented governance or ethical frameworks in the form of guardrails for safety, privacy and bias mitigation; and
Thirty-nine percent cite high implementation cost as the primary barrier to gen AI adoption.

AI maturity requires foundations

“The road map aims to help guide stakeholders to fully understand the scope of gen AI, and to build transparent strategies that deliver on its promise without placing enterprises at risk,” says Goldstuck. “What’s most startling is that many companies think using a gen AI tool is the same as having an AI strategy.”

As companies race to embed gen AI tools like Microsoft Copilot and ChatGPT into business functions, most are overlooking deeper transformation through infrastructure, skills and internal capability. Holistic AI infrastructure, combined with people and processes, is critical to scaling AI deployments and clearly connecting them to tangible return on investment.

Shadow AI

The report raises the alarm about “shadow AI” – the unsanctioned use of gen AI by employees without oversight. Currently:

Thirty-two percent of businesses report informal or unregulated gen AI use;
A further 20% report a mix of official and unofficial gen AI use; and
Eighty-four percent say oversight is an important or very important success factor for gen AI deployment.

Critical governance measures include clear principles for oversight, accountability and responsible use. It enables organisations to build trust, reduce risk and drive long-term value.

“The current use of gen AI is largely taking place in a regulatory and ethical vacuum,” Goldstuck warns. “The longer this continues, the more harm can be caused, to both businesses and individuals, before these guardrails are in place. Without governance, organisations are walking blindfolded into a future shaped by AI. That might be exciting, but it is not sustainable.”

The road map also identifies two areas of opportunity:

Business and societal impact: Over 75% of respondents have no measures in place to monitor or reduce the energy use and footprint of gen AI.
Skills development: A massive 87% of businesses have committed to gen AI upskilling or training of employees.

The report cautions that South Africa could find itself divided by the ability to use gen AI wisely and scale deployments as the technology matures.

Goldstuck says: “There’s a real risk of a gen AI disconnect in South Africa between those who use gen AI deliberately, strategically and ethically, and those who use it blindly or not at all.”