The AI Opportunity

The agreement is real. The opportunity is not automatic

In New York this week, Kenya and Anthropic signed a Joint Declaration on artificial intelligence. The framework covers skills and capacity building, research, responsible public-sector applications, AI safety and evaluation, and possible uses in education and health. It also says future programmes involving data must operate under Kenyan law and respect the country’s data sovereignty.

That is significant. Anthropic develops Claude, one of the leading families of frontier artificial-intelligence models. Kenya is not merely asking how to regulate technology arriving from elsewhere; it is trying to build relationships with companies shaping that technology.

But a declaration signed in New York is not yet an AI economy in Nairobi, Kisumu or Mombasa. It is not a job. It is not a computer in a school. It is not affordable internet. And it does not by itself give a young Kenyan the knowledge, space, confidence or network needed to turn a powerful tool into a useful one.

The more important question therefore begins after the signing: what does an AI opportunity look like before it becomes an AI job?

A small room in Lucky Summer

A useful place to look for an answer is not a conference hall in New York. It is a workspace in Lucky Summer, on the north-eastern side of Nairobi.

M.O.B Hub is deliberately ordinary. It offers desks, computers, internet access, private offices and space for meetings and training. Its website lists a shared workspace at KES 100 an hour, about €0.70 at the Central Bank of Kenya’s 23 September exchange rate, or KES 300 a day, about €2. Computer access and other services add another practical layer to the offer.

Onyango’s own path helps explain why this matters. Growing up in Korogocho, he had opportunities that remain far from ordinary for many young people there: he was able to continue his education and eventually earned a degree in accounting. He did not emerge from university with an investor waiting to finance his next idea. Instead, he took the entrepreneurial risk of building M.O.B Hub himself — turning an opportunity he had received into infrastructure that can create opportunities for others.

Brian Onyango, who built M.O.B Hub, has described that gap with unusual precision: “Digital equity requires more than just an internet connection.” In the same public reflection, he argues that learning rarely becomes productivity if people lack accessible and affordable places in which to use new skills.

During visits to Lucky Summer in January and February 2026, AFRICA3000 observed children and young people from neighbouring Korogocho using the space. That matters because Korogocho is often described from outside through what it lacks. The people using M.O.B are doing something more interesting: working with what is available and expanding what can become possible.

On another visit to Korogocho, Onyango was asked how people there deal with the poverty around them. His answer came immediately: “Poverty is not a category in which we think.”

The sentence does not make poverty disappear. It refuses to make poverty the category through which a person’s intelligence, ambition or future is understood.

That distinction matters for artificial intelligence. The question is not whether a young person growing up in Korogocho has less potential than somebody growing up in Westlands or Karen. The question is whether they have comparable opportunities to discover, develop and use it.

Potential is everywhere. Opportunity is not.

Kenya already has an AI strategy

Kenya’s agreement with Anthropic did not appear in a policy vacuum. The National Artificial Intelligence Strategy 2025–2030, launched in March 2025, is organised around three broad pillars: AI digital infrastructure; data and AI governance; and AI research, innovation and commercialisation.

Two days before the Anthropic announcement, Broadcasting and Telecommunications Principal Secretary Stephen Isaboke told a Nairobi forum organised with the World Bank that Kenya now needs to move from policy to practical results. The government’s own test, he said, should be whether digital investment produces more productive businesses, better jobs, more efficient public services, greater inclusion and improved quality of life.

That is a useful standard precisely because it can be measured against reality. A strategy can create direction. A partnership can create access to expertise. Neither proves that the benefits will spread.

The Joint Declaration with Anthropic is also a framework rather than a finished programme. The next phase is expected to develop specific pilots, institutional partnerships, research initiatives and capacity-building programmes. That makes the details important: who participates, where programmes happen, which languages and sectors they serve, how results are evaluated, and what Kenyan institutions retain after a pilot ends.

AI will not affect every worker in the same way

The loudest global argument about AI and employment often begins with replacement: which jobs will disappear? Kenya’s labour market makes the picture more complicated.

Isaboke cited World Bank analysis indicating that about 87 percent of Kenyan workers are in occupations with relatively low direct exposure to current AI capabilities. About 9 percent are in highly exposed occupations where AI is more likely to complement work. The remaining highly exposed group faces greater displacement risk.

These are estimates of occupational exposure, not forecasts that a fixed percentage of jobs will vanish or be created. They also describe direct exposure. AI can affect workers indirectly when firms reorganise, new services emerge, productivity changes or customers begin expecting different forms of work.

The World Bank’s World Development Report 2026 makes a broader point relevant to Kenya: in low- and middle-income economies, AI may initially complement more workers than it displaces. But the report warns that gains depend on reliable power and connectivity, skills, institutions and locally useful data.

That shifts the question from “Will AI take Kenyan jobs?” to harder questions. Which workers will learn to use it? Which businesses will adapt it? Which students will understand it well enough to question its answers? And who will have the opportunity to build products and services rather than remain only a consumer of systems developed elsewhere?

The AI opportunity begins before the labour market

KENYA | FACTS recently examined the architecture of Kenya’s education system: Competency-Based Education, Senior School, Technical and Vocational Education and Training, universities, funding and the uneven conditions in which learners move through them. The AI story begins inside that same system.

Kenya’s curriculum increasingly emphasises communication, collaboration, critical thinking, creativity, digital literacy and the ability to apply knowledge. Those capacities become more important, not less, when a machine can generate an answer in seconds.

AI can help a learner explain a difficult concept in another way, practise a language, explore programming, compare ideas, summarise material or begin researching a question. It can help a small business draft text, analyse information or test an idea. It can give a young developer access to forms of technical assistance that previously required a colleague with more experience.

But AI also produces errors. It can invent facts, reproduce bias, flatten context and make weak work look polished. The ability to use AI therefore includes the ability not to trust it automatically.

This is why access cannot be reduced to providing accounts for a chatbot. A learner first needs foundational literacy and numeracy. Then comes digital literacy, subject knowledge, critical judgement and enough practical access to experiment. A powerful model does not remove those layers. In many cases, it makes them more important.

Opportunity does not create talent. It gives talent somewhere to go.

Between an online course and a career

M.O.B Hub is interesting because it sits in what might be called the missing middle of digital policy.

At one end are national strategies, fibre networks, data centres and agreements with global technology companies. At the other is the individual learner or worker who is told to acquire digital skills. Between them are the ordinary conditions that determine whether this advice is usable: a device, electricity, connectivity, a desk, time, guidance, peers, mentors and professional networks.

A computer donation can solve one part of that chain. It cannot by itself create the encounters through which people learn how industries work, discover opportunities or meet someone who can answer the next question.

This is also why the ambition for a place such as M.O.B can sensibly extend beyond serving its immediate neighbourhood. Lucky Summer does not have to be treated as a location from which talent must first escape before it can join Nairobi’s professional life. A workspace there can also draw people in — businesses from the Central Business District, researchers, universities, foundations, non-governmental organisations and other groups looking for a place to meet, work or run a workshop.

That possibility matters. A young person from Korogocho and a researcher from Europe do not build a network merely because both have internet access. They build one when their paths cross. Physical spaces can make those crossings less accidental.

The same tool can widen participation — or manufacture it

The Anthropic partnership also arrives with an awkward reminder that AI literacy is not only an employment skill.

In its September 2026 threat-intelligence report, Anthropic described a Kenyan operation in which one actor used Claude to mass-produce political social-media content designed to look like spontaneous grassroots opinion. According to Anthropic, the network amplified messages supportive of government figures and hostile to opposition figures. Anthropic said it found no evidence of government involvement, did not identify the organisation responsible and assessed the operation as having failed to reach real people outside the inauthentic network. The company removed the account and organisation behind it.

The episode should not be inflated into proof that AI has transformed Kenyan politics. Anthropic’s own assessment says the operation had no demonstrated real-world reach. Its importance is narrower and more useful: generative AI can sharply reduce the effort needed to manufacture large quantities of apparently human political speech.

That makes AI literacy a civic skill as well as an economic one. Citizens need to know that fluent language is not evidence of a real person, that repetition is not evidence of public consensus, and that synthetic content can be produced at industrial speed.

Kenya’s AI opportunity therefore includes safety, verification and media literacy. A country cannot become more capable with AI if its citizens become less capable of judging what AI produces.

Using frontier AI without surrendering the local context

One of the most consequential parts of the Kenya–Anthropic declaration concerns data and local relevance. The framework says future programmes involving data are to comply with Kenyan data-protection, cybersecurity and public-sector requirements. It also points toward systems that reflect Kenyan languages, institutions, development priorities and social context.

This is not a technical footnote. The World Bank notes that the most advanced models, chips and data-centre infrastructure are concentrated in a small number of companies and countries. Developing economies can gain enormously by adopting and adapting existing systems rather than attempting to reproduce every layer of the frontier AI industry. But dependence creates its own questions.

Who controls sensitive data? Which Kenyan languages work well enough for serious use? Can local researchers evaluate models rather than simply receive them? Can Kenyan firms build applications on top of global systems and retain meaningful economic value? Can public institutions acquire expertise that remains after an external partner leaves?

The Joint Declaration provides language for addressing some of these questions. The test will come with implementation.

From access to agency

Kenya does not need every young person to become an AI engineer. It needs a society in which AI capability is distributed widely enough that people can decide where the technology is useful, where it is dangerous and where it is irrelevant.

That requires different levels of skill. A teacher needs to know when AI can support learning and when it undermines it. A small entrepreneur needs to know what it can automate without surrendering judgement. A journalist needs to verify synthetic material. A public servant needs to understand the consequences of feeding sensitive information into a model. Developers and researchers need much deeper technical capability to adapt, evaluate and build systems.

The national opportunity is therefore larger than training people to prompt a chatbot. It is about creating the conditions in which Kenyan institutions, companies and citizens can use advanced systems on their own terms.

That brings the story back to Lucky Summer. M.O.B Hub is tiny beside a national AI strategy. It should not be romanticised into something larger than it is. But it illustrates a basic principle that national strategies can easily miss: opportunity becomes real in places.

A desk. A computer. A connection. Someone to ask. Someone to meet. Enough time to learn. A reason to return tomorrow.

Frontier technology may be built elsewhere. Whether it becomes a Kenyan opportunity will be decided much closer to home.

AI development is global; access to its benefits is not

European governments, universities, foundations and companies are already deeply involved in African education, digital infrastructure, development finance and research. AI adds a new layer to those relationships.

The easy model is to bring expertise to Kenya. The more interesting model is reciprocal: work with institutions and people already building useful infrastructure, fund capability rather than dependency, and create professional networks that do not require every promising young Kenyan to leave his or her neighbourhood before being taken seriously.

There is also a lesson for Europe itself. The digital divide is not only a problem of broadband coverage. It can exist inside a connected city, between people who have a device and people who have a useful place to work; between people who can ask an AI system a question and people who know how to judge the answer; between access to information and access to networks.

Brian Onyango’s formulation is therefore larger than M.O.B Hub: digital equity requires more than a connection. The AI era may make that distinction impossible to ignore.

The agreement now has to become specific

The pilots: Which ministries, universities, schools, health institutions or research organisations become the first partners under the Kenya–Anthropic framework?

The skills programmes: Whether capacity building reaches beyond public officials and established technology institutions to teachers, students, small businesses and communities with fewer digital resources.

The data rules: How Kenya operationalises data sovereignty, privacy and cybersecurity when advanced AI systems are used in public-sector settings.

The jobs evidence: Whether Kenya begins publishing regular, occupation-level evidence on where AI is complementing workers, changing tasks or creating displacement pressure.

The local ecosystem: Whether Kenyan developers, researchers and firms gain the resources to adapt, evaluate and commercialise AI rather than only consume it.

The physical layer: Whether investment in digital skills is matched by affordable places where people can actually learn, work, collaborate and build.