AI could add as much as two percentage points to South Africa’s annual economic growth rate, on top of whatever the economy would otherwise manage, according to Efficient Group chief economist Dawie Roodt – an estimate that towers over what either the World Bank or PwC have put on the table.

“AI can easily add 2% to GDP growth,” Roodt has told TechCentral, responding to the World Bank’s World Development Report 2026, which calls AI a lifeline for developing economies and finds that they have more to gain – and less to fear – from AI than richer ones, because far fewer of their jobs are exposed to automation. That is a claim about the balance of risk and reward, not about the size of the prize: the bank’s own modelling still puts the productivity dividend to emerging markets at well under half that of advanced economies.

Asked whether he meant the annual growth rate or a one-off lift to the level of output, Roodt confirmed the former: whatever growth turns out to be, AI can add 2% on top of it, each year. The lower end of his range, 1%, is what he expects as things stand; the 2% assumes the skills and policy constraints are dealt with.

“Current GDP growth is less than 1%, for most of the past 15 years,” Roodt said. Population growth of about 1.4% on his estimate means “we are getting poorer on a per capita basis” and that “most GDP growth comes from population growth”.

Stats SA’s latest mid-year estimates put population growth lower, at 1.2%, but his broader point holds: the economy grew 1.1% in 2025, just short of the rate at which the population expanded.

Growth picked up to 1.9% year on year in the first quarter of 2026. Roodt reads the longer-run shortfall as a productivity failure: “Under normal circumstances, productivity growth should add approximately 1%, which means our productivity growth is dismal.”

PwC much less bullish

PwC’s modelling, published in its South Africa Economic Outlook in October last year, makes a very different finding. Under what it calls a tense transition – “probably the most realistic future scenario”, in which regionalisation and nationalism fragment the technology landscape – South Africa sees “a small annual gain in real GDP and an aggregate benefit of 1.2 percentage points over the 10-year period” to 2035.

That is cumulative, not annual, and PwC is explicit about the difference. Globally it puts the upside at up to 15 percentage points of output over the decade, which “would effectively add one percentage point to annual real GDP growth rates”. On the same basis, 1.2 points over 10 years works out to roughly a tenth of a point a year – making Roodt’s estimate more than 15 times PwC’s most realistic local scenario, and about two-and-a-half times the World Bank’s most optimistic figure for emerging markets.

The gap narrows if both sides are read as conditional. PwC’s better outcome, if South Africa “can reduce inequality of access to AI tools”, is six percentage points towards 2035 – still cumulative, or about 0.6 of a point a year. Roodt’s conditional 2% is more than three times even that. In a “turbulent times” scenario, where AI destroys more work than it creates, PwC has it subtracting from GDP altogether.

The reason for the local discount is structural: the country “has a smaller potential economic benefit from AI compared to the global average due to high levels of social inequality”, with a large marginalised and informal sector that has almost no exposure to the technology. Only 2% of South African job vacancies in 2024 specified AI-related skills, about 17 000 of 845 000.

“Like electricity in the 20th century, AI has the potential to create more jobs than it displaces if it is used to pioneer new forms of economic activity,” PwC South Africa chief economist Lullu Krugel says in the report. “But – and this is very important – South Africa needs to use AI to the benefit of society as a whole, or risk leaving behind many people that have less access to technology.”

The World Bank’s view

The World Bank’s numbers fall between the two. Extending a method devised by MIT economist and Nobel laureate Daron Acemoglu across 52 countries, it finds expected productivity gains in advanced economies are “more than double” those in emerging markets: its most optimistic scenario lifts advanced-economy potential growth from 1.2% to 3.6% over the 2020s, against 4.1% to 4.9% for emerging markets, or eight-tenths of a point.

Chief economist Indermit Gill nonetheless calls AI a lifeline “they should grasp before it slips away”, arguing that even the worst case would lift developing-economy potential growth “above the dismal average of the first half of the 2020s”, at a time when development progress “has dipped to a 75-year low”. The mechanism is scarcity: “A technology tends to achieve the largest gains where the resource that it augments is in shortest supply.”

Only 4.5% of jobs in low- and middle-income countries are susceptible to automation by generative AI, against 14.2% in high-income countries. Gill’s “more to gain” rests on that ratio rather than on the growth projections. Where they agree, the two groups are close: 16.2% of jobs in developing economies could see a meaningful productivity boost, against 18.7% in high-income ones. The bank warns the adoption gap between rich and poor countries is set to widen.

Skills, infrastructure and policy

Roodt’s conditions are not technological. “Skills is a major constraint. And infrastructure. And policy,” he said, pointing to the delay in licensing satellite broadband, which has kept SpaceX’s Starlink out of South Africa, and to the regulatory settlement around mining, a reference to mineral & petroleum resources minister Gwede Mantashe and black ownership requirements.

He also expects the gains to come from an unexpected place – “the primary and secondary part of the economy is likely to benefit most” – though the World Bank puts agriculture and manufacturing employment among the least exposed to AI, and PwC’s index ranks labourers and machine operators at the bottom of its own.

Read: SA experts split on whether the singularity has begun

All three accounts converge on uptake as the binding constraint. About two-thirds of the productivity gap between rich and developing economies is not structural on the bank’s calculation, but a matter of adoption.

By April 2025, nearly 25% of internet users in high-income countries had used ChatGPT, against 5.8% in upper-middle-income countries, the band South Africa falls into, and the bank projects the adoption gap widening from 24 percentage points to 46pps by 2035. “The window to get this right is narrow,” said Gaurav Nayyar, who directed the report.

South Africa may be poorly placed to hit it. The bank counts more than 80 countries with published national AI strategies as of mid-2026 and excludes South Africa because communications minister Solly Malatsi withdrew the draft national AI policy on 26 April, after the 86-page document was found to contain fabricated academic citations.

A redrafted, lighter-touch version is due at cabinet in November. The same report names South Africa’s “chronic load shedding” (now seemingly resolved) as a reason hyperscale AI investment is going to Brazil and Malaysia instead, and pegs local nominal GDP at about $427-billion against roughly $775-billion in projected 2026 capital spending by five American AI hyperscalers.

Roodt’s arithmetic survives all of this, but only as a conditional. “Without AI, certain policy changes can lift GDP to 3% easily,” he said. “Add say 2% of AI to that, then a 5% GDP is certainly possible, even more.”