Despite widespread artificial intelligence (AI) adoption, most large South African organisations are struggling to determine whether the technology is delivering measurable financial returns, with many unable to quantify its contribution to cost savings.
According to The SA AI Index 2026, 73% of large organisations have no tools to assess whether their AI deployments deliver value, while 71% cannot put a figure on the cost savings generated by the technology.
The study, conducted by World Wide Worx, in partnership with Altron Digital Business, reveals a significant gap between AI implementation and the ability to establish its business impact. It is the third edition of the annual research into AI adoption in SA.
The findings suggest that while businesses are increasingly incorporating AI into their operations, measurement of the return on investment remains a major weakness. The value pillar of the newly-introduced SA AI Index scores SA at 16.6 out of 100, the lowest result across its five measurement categories.
This low score suggests the country is struggling to turn AI capabilities and investments into real results.
Arthur Goldstuck, CEO of World Wide Worx, says the growing prevalence of AI across SA has not necessarily translated into meaningful business improvements.
“AI has gone from novelty to normal in South African business in just two years,” he notes.
“But using AI and getting value from it are very different things. The early adopters were enthusiasts who went looking for gains. Today, almost everyone uses AI, often in a shallow way, and many organisations are discovering that a language model on every desk does not automatically translate into business results.”
The research finds that organisations deploying AI in live business processes report stronger productivity gains than those still experimenting.
The SA AI Index 2026 research was conducted using telephonic interviews with 100 decision-makers from South African organisations employing more than 200 people.
The index measures how effectively organisations are integrating AI into their operations and translating its use into measurable business benefits, with results scored out of 100.
Measurable impact
The SA AI Index measures how deeply organisations have integrated AI into their operations, rather than simply whether they use the technology. It considers five pillars: deployment, adoption, workforce, governance and value.
The average large South African organisation scored 34.3 out of 100, indicating that most businesses have made limited progress towards embedding AI across their operations.
Deployment recorded the highest score at 43.8, followed by adoption at 41.8, workforce at 37.2 and governance at 32.
However, the significantly lower value score highlights the difficulty organisations face in determining whether their AI investments are translating into tangible business outcomes.
The research also points to a difference in outcomes, depending on how extensively organisations have implemented the technology. Among the 30% of businesses with AI operating in a live business process, 80% report a positive effect on productivity.
Thiscompares with 44% of organisations still experimenting with AI and just 6% of those using it on an ad hoc basis. More than half (52%) of organisations remain at the experimental stage.
"The real productivity gains from AI emerge when organisations move beyond experimentation and embed the technology into their day-to-day business operations," the study finds.
From deployment to value
Craig Stewart, MD of Altron Digital Business, says businesses need to move beyond isolated AI applications and incorporate the technology into their core operations.
“Every boardroom is talking about AI, and the interest is real,” says Stewart. “But this research shows the gap between having AI and having it at the heart of the business. The organisations seeing real gains have taken AI into production, built it into their processes and put measures in place to track what it returns. The next wave of value will come from this arena.”
The study also introduces the SA AI Frontier Index, which measures the extent to which organisations use AI to develop new capabilities and change existing business operations, rather than simply improve the speed of established tasks.
The average organisation scored 20.3 out of 100 on this measure, while only 8% achieved a score of 50 or higher.
Only 17% have AI operating in business functions that change what an organisation produces or how it operates. These applications include developing new products, software development, quality control and supply chain management.
Furthermore, just 14% have configured or fine-tuned an AI model for their own requirements, while 7% have developed their own models.
Stewart says the next stage of AI development will require organisations to expand their ambitions beyond productivity improvements.
“The AI Frontier Index shows where the next opportunity lies,” says Stewart. “Most organisations are using AI to make existing work faster, and that has real value. But the organisations that will pull ahead are those using AI to create new products and services, run their operations and supply chains differently, and build AI capability of their own. Very few South African organisations have reached that frontier yet, which means the opportunity is still wide open.”
The research further shows the depth of AI integration is influenced by organisational size, with larger businesses recording stronger results.
Organisations employing more than 1 000 people achieved an average SA AI Index score of 42.6, compared with 33 for those employing between 501 and 1 000 people and 28 for organisations with between 201 and 500 employees.
Only 13% of organisations achieved a score of 60 or higher, indicating that relatively few have made substantial progress in integrating AI across their businesses.
Goldstuck explains that the index provides a baseline against which South African organisations can track their progress as AI adoption matures.
“The SA AI Index gives South African organisations a benchmark for the first time,” says Goldstuck. “A score by itself means little, but we can now measure movement off the 2026 baseline.”
The full report is scheduled to be made available online on 23 November.

