AI Engineer World's Fair 2025

Unlocking Africa's Potential with AI — Thabang Ledwaba

Loading slides…

Read the talk

Unlocking Africa’s Potential with AI

From mobile money to identity applications, Thabang Ledwaba explores how Africa can use AI to solve local problems and become a producer of technology.

From a talk by Thabang Ledwaba

What does AI mean for Africa?

What does AI mean when the conversation moves beyond technology meetups and into everyday life? For Africa, the question is whether it can help transform the continent—and what that transformation would actually involve. The rapid succession of terms can make the opportunity difficult to distinguish from the excitement surrounding it.

Thabang Ledwaba opens with a broad historical sketch. His slide groups neural networks under 1950–1970, machine learning under 1980–2010, and deep learning under “Today.” It is a simplified picture of successive waves of attention. The next shift changes the question again: generative AI supplies outputs, while agentic AI raises the possibility of systems making decisions for us. That expanding scope helps explain why AI can feel both promising and overwhelming.

History of AI slide with three icons labeled Neural Networks 1950–1970, Machine Learning 1980–2010, and Deep Learning Today above a horizontal arrow.Open full source frame
The presentation’s AI history timeline: neural networks, machine learning, and deep learning.
0:120:24
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

0:12 · section reference included

A name, several interpretations

At a conference in the Netherlands, Ledwaba took part in discussions about what counted as AI. Some websites advertised as AI-powered did not withstand that scrutiny. As the disagreements developed, participants reverted to their home languages. He found himself outside a conversation he had helped provoke.

The organizers tried to make amends with a visit to a famous pub. There, introducing himself produced an unexpected response: a beer. His name sounded close enough to a familiar Dutch beer for the introduction to be interpreted as an order. Similar misunderstandings happen in Afrikaans-speaking communities, where saying his name quickly can prompt questions about whether he is frightened. In Setswana, however, Thabang means “to be happy.” As his father’s firstborn son, he jokes, perhaps he was evidence of some happiness beneath the worries of parenthood.

1:411:57
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

1:41 · section reference included

From taking devices apart to studying systems

Growing up in Soweto, Johannesburg, Ledwaba liked breaking objects into their constituent pieces. His father’s belt-mounted pager was an early target: why carry something so heavy, and did its weight explain the distinctive way pager owners walked? Older mobile phones posed another puzzle—how could this one object make a call? The bulky television at his grandmother’s home was more mysterious still. Why did it need so much space at the back? Were the actors inside? His investigations had lasting consequences: he says he is still buying televisions for his grandmother and family.

Medicine initially offered another way to investigate how things worked, although repairing the whole mattered more there than dismantling it. His attention gravitated toward the equipment: how did sensors detect a patient’s condition, what produced a flatline, and how did life-support systems keep someone alive? That curiosity helped take him from Wits, the University of the Witwatersrand, to the University of Johannesburg, where he completed a BSc in computer science.

The same habit of examining components suggested a larger question. Before treating Africa as a single story about development or disadvantage, what capabilities become visible when its parts are examined? What does the continent offer the world beyond the natural minerals for which it is so often recognized?

3:484:03
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

3:48 · section reference included

Adoption is already happening

Africa is often portrayed as a newcomer still learning what AI means. Ledwaba reports reading that Kenya ranked third for daily ChatGPT use, although he does not identify the ranking’s date or population. His broader point is that people are already using these tools. The harder question concerns invention: when he looks at AI patents and new startups, Europe and Asia seem much more visible. Using technology and shaping its development are different forms of participation.

Limited resources can encourage a useful discipline: simplicity. M-PESA provides a concrete precedent in fintech, allowing people to transfer money using a recipient’s mobile phone number. The significance lies in solving a practical coordination problem with technology people can access. Ledwaba’s comparison of fintech revenue in 2021 with projections for 2030 presents Africa as a potential global participant in that market.

Fintech is not automatically AI. A useful digital service does not become an AI implementation merely because it introduces a new tool or process. M-PESA establishes a precedent for solving local problems with technology; it does not establish that those solutions require AI. With companies in South Africa already specializing in AI, the next question is what that additional capability can contribute.

Ledwaba cites an unnamed forecast of roughly seventeenfold growth in an African market-size and investment discussion from 2020 to 2030, correcting his initial “seventeen percent.” He describes the chart in billions of dollars, but does not specify the market definition or baseline. The distinction that matters for the rest of the talk is between a growing technology market and a growing capacity to improve people’s lives. Investment can indicate commercial interest without answering whether the resulting services address the continent’s problems.

7:057:20
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

7:05 · section reference included

Improve the queue—or reconsider the application?

Losing an identity document creates two problems: the missing document and the prospect of returning to Home Affairs. For Ledwaba, the dread of standing in that queue makes the gap between technology and impact tangible. Politics, tax laws, finances and administrative rules can constrain changes to the service itself. Teams then concentrate on improving the process they are allowed to touch.

Several responses are reasonable on their own:

ResponseWhat it changes
Ticketing systemHow the queue is organized
More staffHow many people can be served
More officesWhere people can obtain service
Online appointmentsWhen people are asked to arrive

These options address different parts of the visit, but all retain the visit as the center of the process. Ledwaba acknowledges their usefulness while asking whether they remove the underlying burden.

Despite online booking, Ledwaba describes Home Affairs queues as roughly twice as bad as five years earlier, a personal observation rather than a measured service comparison. Discussions with delegates continued through ticketing systems and procurement tenders. Then he asked a question upstream of all those proposals: why must a person start an application when the state already knows when they were born and the age at which they qualify?

Known eligibility could trigger an application instead of waiting for a citizen to initiate it. The proposed sequence is straightforward:

  1. Use the recorded birth date.
  2. Determine when the person reaches the qualifying age for the document.
  3. Initiate the application process when that condition is met.

Ledwaba praises the newer South African ID design, but distinguishes improving the document from improving the process that produces it. His proposal concerns automatic initiation; it is not a demonstration of automatic approval or issuance.

For a concrete illustration, consider a person whose recorded birth date now satisfies the eligibility rule but who has not opened an application. Before the proposed trigger, the birth record exists and the application does not. Afterward, the birth record is unchanged and an application has been initiated; the document has not thereby been issued. This makes the change in responsibility visible: the service starts work from information it already holds.

Proposed eligibility-triggered application

Constructed example: The identifier person-001 and record labels are teaching details. No birth date, statutory qualifying age, deployed workflow or completed issuance is asserted.

Eligibility evaluation for person-001 — unchanged
Recorded birth date satisfies the qualifying-age rule; no application has been opened.

Operation: On reaching eligibility, initiate the application process using the existing birth record.

person-001 birth record

Before: Before the proposed trigger
Birth date recorded
After: After the proposed trigger · Unchanged
Birth date recorded

person-001 eligibility

Before: Before the proposed trigger
Qualifying age reached
After: After the proposed trigger · Unchanged
Qualifying age reached

person-001 application

Before: Before the proposed trigger
Not present
After: After the proposed trigger · Added
Initiated

person-001 issued document

Before: Before the proposed trigger
Not present
After: After the proposed trigger · Unchanged
Not present
The service begins the application from known eligibility; issuance remains a separate outcome.

The potential relevance extends beyond South Africa to identity documents, passports and other essential paperwork. The engineering insight is the change in starting point: examine why the process begins with a user request before investing further in managing the requests that arrive. That question can expose a simpler intervention than another layer of queue management.

11:2811:44
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

11:28 · section reference included

Local constraints can produce widely useful ideas

Other examples extend the discussion from administration to access to essential goods:

  • Medicine delivery: Ledwaba describes PharmaWings as delivering medicine to remote or difficult-to-reach regions of Nigeria.
  • Drinking water: He describes Project Cumulus as producing drinking water from air, tentatively locating it in Morocco.

These are examples he offers of practical invention, without establishing their AI components. Their relevance is the starting problem: reaching people and supplying something they need.

The geographical scope matters, too. Ledwaba begins to describe South Africa’s achievements, then broadens the statement to Africa. Creativity, technological use and opportunity are already present across the continent. The question becomes how to turn those capabilities into improvements large enough to change existing circumstances.

16:3916:59
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

16:39 · section reference included

Recognize capability, then retain more of its value

One obstacle, in Ledwaba’s view, is how Africans assess their own capabilities. Seeing oneself as below par can make matching other regions seem like the only available ambition. That self-image sits awkwardly beside international investment and the recruitment of innovators. He points to Chinese automotive businesses recruiting innovators as an example of others recognizing capabilities that may be undervalued at home.

Natural resources make the question of value especially visible. Ledwaba’s figure of about 30% corresponds to the historical estimate of Africa’s share of global mineral reserves used by UNEP in 2016, not a share specifically of rare-earth elements. During a visit near Victoria Falls, a conversation with a chief turned to the continent’s mineral wealth and its exports. The accompanying map locates resources including bauxite, cobalt, copper, graphite, lithium, manganese, nickel and platinum across African countries.

Map of Africa with country labels and colored dots keyed to bauxite, cobalt, copper, graphite, lithium, manganese, nickel, and platinum.Open full source frame
A map of Africa marks mineral resources across countries.

Exporting minerals produces some income. But processing and manufacturing turn those inputs into products such as car parts and cars, which African countries then buy back. The concern is where the transformation happens and who benefits from it. Possessing an input does not mean capturing the value of the finished product. That sequence leads Ledwaba to ask whether the continent recognizes its own potential and the power of acting together.

17:4318:06
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

17:43 · section reference included

Build for local needs, participate as producers

Ledwaba presents China as an example of changing a country’s position by serving domestic needs before international markets. The principle he draws from that comparison is to make solutions work for the people who need them locally. Some of those problems are shared elsewhere, creating a path from local usefulness to global relevance. Creativity and access to AI may shorten the time needed to develop such solutions, even under resource constraints.

That also changes what meaningful international support looks like. Financial investment alone does not answer whether African builders have the opportunity to participate in setting direction and producing technology. Ledwaba asks what the rest of the world is doing to grow Africa’s potential and give its innovators a place where their work can be seen.

Ubuntu and simplicity supply a basis for that work: a sense of shared possibility, coupled with attention to what is actually needed. Questions about a united Africa or a common currency come after a more basic decision about the goal. The ambition is to participate as producers, not only as consumers. Without that goal, debates about the institutional form risk taking precedence over the outcome they are meant to enable.

20:2220:33
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

20:22 · section reference included

Use the tools and define the opportunity

There is already a foundation to build on: companies investing in AI, universities teaching it, and primary and high schools beginning to discuss its possibilities. Ledwaba also describes a projected increase of about 72 million rand in the South African intelligence market over the following ten years; the forecast’s source, baseline and precise market definition are unspecified. These signs of activity return him to the question of position: are Africans treating themselves as late arrivals even while they are already participating?

The practical response is to move beyond the large words surrounding AI and experiment with the tools. For an African builder, the opening question—what does AI mean to you?—can now be made specific: which existing problem can you approach differently with the resources and creativity available? For a European audience, the corresponding question is what would help Africa stand on its own feet as an active market participant.

AI may help people think through situations, but its role need not be to supply every capability that was missing. It may instead make existing capabilities easier to recognize and use. The opportunity is therefore partly technological and partly a decision about what to attempt. When Ledwaba invokes the world as an oyster, its meaning depends on the person choosing what to build.

Thabang Ledwaba identifies himself as a senior software engineer and R&D engineer at BBD Software.

23:0623:28
Suggest correction

This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.

23:06 · section reference included

Resources

Transcript download
Full transcript

Loading transcript…