Asher's Blog

Africa has heard the language of leapfrogging before. Mobile phones would allow us to bypass fixed telephone infrastructure. Mobile money would allow people excluded from banking to participate in finance. Cloud software would reduce the need for companies to build expensive local technology. Some of those promises became real, but many institutions still remained manual, fragmented and difficult to navigate.

Artificial intelligence creates another leapfrogging opportunity, and this time the opportunity reaches beyond access to a single service. AI may allow emerging markets to compress the cost and time required to build capabilities that previously depended on generations of institutional knowledge.

This does not mean Africa can skip institutions themselves. Trust, law, accountability and public infrastructure still matter. The opportunity is to skip some of the slow, paper-heavy and labour-intensive methods through which institutions traditionally developed.

Intelligence can become infrastructure

In many developed economies, a business can draw on layers of institutional support: searchable public records, mature credit systems, professional services, standardised data, reliable addresses and experienced administrators. African entrepreneurs often rebuild pieces of this infrastructure inside their own companies because the surrounding system is incomplete.

AI can reduce that burden. A small business may receive help interpreting regulation, organising records, forecasting cash flow and responding to customers without employing a specialist for every function. A public institution may digitise and classify documents faster, make information easier to retrieve and support employees handling large volumes of requests.

The African Union’s Continental AI Strategy recognises this opportunity while calling for an Africa-centred, development-focused and responsible approach. The significance is not simply that governments want more AI companies. It is that intelligence can become a layer across agriculture, education, health, financial services and public administration.

Leapfrogging begins with the workflows we actually have

Africa will not gain the full advantage by importing systems designed around foreign institutions and placing a chatbot on top. Our workflows, languages, informal economies, infrastructure and risk environments are different.

A useful AI system for a Nigerian small business must understand that records may be fragmented across bank statements, spreadsheets, paper and WhatsApp conversations. It must work within connectivity and device constraints and recognise the regulatory environment in which the business operates. A system for agriculture must understand local crops, weather, market access and languages rather than merely translate advice created for another geography.

The opportunity is to design from present reality towards a better system. If we first force every institution to resemble the systems of the West before applying intelligence, we will reproduce their historical sequence. Leapfrogging means asking what outcome the institution exists to create and whether new tools allow us to reach it through a different architecture.

The missing institution cannot become the missing accountability

AI can help deliver institutional capability, but it cannot be allowed to hide the person or organisation responsible for a decision. A citizen denied a service, a borrower rejected for credit or an employee affected by an automated process must have a path to explanation and appeal.

Weak institutions sometimes rely on discretion without records. AI could improve consistency and auditability, but it could also scale invisible bias or make poor decisions appear scientific. The same technology that reduces administrative delay can make errors harder to challenge if leaders treat the model as the final authority.

Leapfrogging must therefore include governance. We need standards for data, privacy, procurement, security, human oversight and redress. Skipping paper should not mean skipping due process.

Local data is part of sovereignty

AI trained mainly on information from outside Africa will misunderstand some of our language, context and priorities. The answer is not to reject global models, but to build the data, evaluation and domain knowledge that make them useful here.

Governments and companies should treat clean local data as development infrastructure. Health records, business registries, educational material, agricultural information and African-language resources can improve services when governed properly. Without them, countries will consume intelligence built from other societies while remaining unable to shape how their own realities are represented.

This creates an important responsibility. Data should not be extracted from citizens and communities without protection or shared value. The institutions collecting it need clear purposes, security and rules that prevent convenience from becoming surveillance.

The leapfrog will be uneven

AI still depends on electricity, connectivity, compute, devices and skilled people. GSMA has highlighted Africa’s compute-capacity gap even as smartphone adoption and mobile infrastructure create new possibilities. A cloud model that is unaffordable, too slow or unavailable during an outage cannot become dependable infrastructure.

This is why edge AI, smaller models and systems designed for intermittent connectivity matter. Africa does not need only larger data centres; it needs intelligent products that respect the economics of the people using them.

The divide may also widen between organisations that can organise their data and those that remain trapped in unstructured records. Large institutions may adopt AI first while small businesses receive generic tools that do not understand their work. Policy should therefore support shared infrastructure, open standards, local research and practical adoption rather than focusing only on headline projects.

We can compress learning, not eliminate it

AI gives a young professional access to explanations, examples and assistance that previously required years inside a mature organisation. It can help a new company adopt financial controls, draft operating procedures and understand global standards earlier. This can compress institutional learning.

However, knowledge delivered by a model is not the same as judgment formed through responsibility. Institutions still need people who understand consequences, challenge the output and adapt it to reality. We can shorten the path to competence, but we cannot eliminate the work of building trust.

The real leapfrog is not a future in which Africa avoids institution-building. It is one in which we build institutions that are intelligent from the beginning: digital records rather than inaccessible files, continuous controls rather than retrospective discovery, services organised around citizens rather than departments and knowledge that survives the individual employee.

This time requires deliberate builders

The opportunity will not materialise because AI exists. It will require African founders, public servants, universities, investors and regulators to choose important problems and build for the conditions around them.

We should focus on places where institutional weakness imposes daily costs: business registration, justice, healthcare access, compliance, education, agriculture, identity and public finance. We should measure success by improved outcomes rather than by the number of AI announcements.

Previous technology waves allowed Africa to skip pieces of physical infrastructure. AI may allow us to skip parts of the administrative and knowledge infrastructure that made development slow. But it will work only if we combine intelligence with local data, responsibility and courageous implementation.

The leapfrog is real this time. The question is whether we will use it to build stronger institutions—or merely faster versions of the weak systems we already have.