South Africa’s traditional measure of the digital divide — whether people can get online — is no longer adequate, Deputy Minister in the Presidency for Women, Youth and Persons with Disabilities Mmapaseka Steve Letsike told the first Youth, AI and the Future of Work Summit, setting out a three-tier framework in its place.
Speaking at Wits University at an event hosted by the South African Digital Governance Forum under the theme “The AI Divide – Ensuring Youth, Women and Persons with Disabilities are Not Left Behind,” Letsike distinguished between connection, capability and command.
“Connection gets a person through the door, capability determines what they are able to do once they enter,” she said. “Command is something deeper still. It is the ability to shape the technologies themselves; to build the enterprise, understand the algorithm, govern the data, own the intellectual property, participate in setting the rules and share meaningfully in the wealth that these technologies create.”
Why the third tier matters
The framework is useful because almost all African AI inclusion work operates at the first two levels.
South Africa’s skills programmes are substantial. Microsoft reports engaging four million learners, training 1.4 million people and issuing credentials to nearly 500,000 citizens, with the SABC partnership extending that further. Absa’s ElevateHer has trained more than 10,000 learners across nine African markets. SA Connect has expanded broadband reach.
Almost none of that touches command. Training someone to use AI tools does not give them a stake in the models, the data governance or the intellectual property.
“A household can have Wi-Fi and still remain economically excluded,” Letsike said. “A young person can own a smartphone and still lack the literacy required to understand the systems shaping their digital environment.”
The digital native fallacy
Letsike was direct about a common assumption. “Knowing how to scroll through TikTok does not necessarily mean understanding why TikTok selected the next video.”
That has empirical support. Research covered by iAfrica found the term “AI” barely registering for 73% of South African respondents, and PwC found that while 64% of African workers had used AI in the past year, only 17% use AI agents daily — high surface adoption, shallow understanding.
Letsike called for greater investment in AI literacy alongside access to finance, procurement and markets for innovators historically excluded from them. Finance and procurement are command-tier interventions; literacy alone is not.
Bias with a gender
On women, Letsike acknowledged AI’s potential to lower barriers to entrepreneurship, expand access to markets and make flexible learning easier — then named the condition.
“Automation can inherit a gender,” she said, warning that existing inequalities can be reproduced through AI systems and biased data. Women, she argued, must be represented among those developing technologies, financing innovation, governing platforms and owning the wealth generated.
That risk is documented in African contexts. Research on business process outsourcing found women facing roughly 10% higher automation exposure because of their concentration in junior roles.
Accessibility as architecture
Letsike’s sharpest design argument concerned disability. AI offers real gains — speech recognition, image description, predictive text, intelligent interfaces — but technology built without considering disabled users creates new exclusions.
“Accessibility must therefore cease to be regarded as an accommodation added at the end of innovation,” she said. “It must become part of the architecture of innovation itself.”
That principle is beginning to appear in African AI funding. The HAIDI Innovation Track, offering grants up to five million Kenyan shillings, requires innovators to work directly with disability communities on testing and validation rather than designing for them at a distance. Kenya’s iHUB and Mastercard Foundation cohort selected six ventures building assistive technology, including Signvrse on sign language translation and DEAFHEALTH. South African manufacturer Lesco has automated production while retaining differently abled workers.
Command without a framework
Letsike urged Africa to develop its own technological capabilities rather than becoming a consumer of technologies developed elsewhere — the argument now common across the continent’s AI discourse.
It arrives while South Africa still has no national AI policy. The draft was withdrawn in April over fabricated citations, with a redraft due at Cabinet in November and public consultation targeted for January 2027. The command tier Letsike describes — who governs data, who owns intellectual property, who sets the rules — is exactly what a national framework would define.
“The frontier now before us asks whether they will have the agency to shape what comes next,” she said.
Based on reporting by SAnews.gov.za.
