advertisement
South Africa’s AI Boom Has A Measurement Problem
Cyril Ramaphosa, South Africa’s President, spent this week in California selling the country as Africa’s home for artificial intelligence. Back home, a survey of the country’s largest employers has raised a harder question: almost all are using AI, but few can say whether it is paying off.
The SA AI Index 2026, produced by research house World Wide Worx with Altron Digital Business, found that 93 per cent of organisations with more than 200 employees now use generative AI in some form. That is up from 45 per cent in 2024 and 67 per cent last year. Adoption is approaching universal levels.
Evidence of value is harder to find. Only 27 per cent of the organisations surveyed can measure AI’s value. Nearly three in four have no tools to assess its contribution, while 71 per cent cannot quantify cost savings. Just over half of users, 51 per cent, rate AI’s effect on productivity as positive.
advertisement
“That’s the scary number for a lot of organisations,” said Arthur Goldstuck, Chief Executive of World Wide Worx, referring to the measurement gap.
The survey drew on telephone interviews with 100 decision-makers between September and November. Its authors caution that findings for smaller groups are based on limited numbers. Even so, the differences between organisations at various stages of adoption are striking.
Among the 30 per cent of organisations that have AI running in a live business process, 80 per cent report a positive effect on productivity. Among the 52 per cent still experimenting, that figure falls to 44 per cent. Companies able to measure AI’s value are also about twice as likely to report a benefit as those that cannot.
advertisement
Craig Stewart, Managing Director of Altron Digital Business, said the findings pointed to “very much a leadership and a cultural challenge”.
The index found that three in four organisations have employees using AI tools that their companies have not approved. Only 1 per cent actively use AI agents, software capable of carrying out tasks autonomously, while 71 per cent have no plans to adopt them. Financial services led the sectors surveyed, followed by telecommunications, while the public sector scored poorly.
The findings arrived as Ramaphosa led a delegation of nearly 300 ministers, executives, researchers and students through Silicon Valley, on a visit running until 21 November.
advertisement
Speaking at Semafor’s Silicon Valley and the World event on Monday, Ramaphosa said South Africa hosts nearly 70 per cent of the data centres built on the African continent. He also said he planned to meet representatives of OpenAI, Anthropic, Microsoft and Google, as well as their investors.
The itinerary includes Stanford University and a workshop in San Jose on AI infrastructure, government applications, safety and the future of work. The delegation is also visiting Alphabet, Cisco, Palo Alto Networks and Microsoft.
At the same time, the government is pursuing a second line of international cooperation. In Wuhan last week, Nomalungelo Gina, Deputy Minister of Science, Technology and Innovation, met Zhang Guangjun, China’s Vice Minister of Science and Technology. The two advanced plans for an AI talent development and innovation centre in South Africa, intended to address the country’s digital skills gap.
“The centre will strengthen innovation capabilities,” Gina said. A Chinese technical team is expected to visit South Africa for a preliminary assessment, according to her department.
The two engagements illustrate the government’s effort to develop South Africa’s AI economy through international investment, computing infrastructure and skills partnerships. It is courting American technology companies and capital while pursuing cooperation with China on talent development.
But the survey highlights what must happen between those partnerships and any economic return: businesses need to integrate AI into their operations, establish oversight and measure its contribution.
Goldstuck’s findings suggest that organisations are more likely to report productivity gains when they move beyond experimentation and deploy AI in live business processes. Those that measure its value are also better placed to determine whether their investments are working.
For the rest, adoption risks becoming a statistic without a business case. They have put AI into employees’ hands, but have yet to establish what it is delivering.
The full report is due on 23 November.