Reaping the Rewards of the Next Technological Revolution: How Africa Can Accelerate AI Adoption Today

Artificial intelligence (AI) could contribute up to $15.7 trillion to the global economy by 2030, making it one of the biggest economic opportunities available to countries and their leaders. Africa has made some credible strides to develop its AI ecosystem through startup acts and research hubs, and has started to leverage AI to improve its public services in the following ways:

1. Food security and agricultural output: Spatial information from flying (drone) sensors has been used to analyse and manage crop health in Kenya and Mozambique, while drought-forecasting tools are being employed in South Africa to predict these events more reliably.

2. Health-care delivery: AI chatbots have been used to triage scarce health-care resources in Rwanda while AI is being used for the drone-based delivery of blood in both Rwanda and Ghana.

3. Improved government services: The government of Togo used AI to analyse satellite imagery to identify the country’s most vulnerable citizens to enable more effective delivery of economic aid during the Covid-19 pandemic.

4. Better communications and access to public resources: Natural-language processing and translation tools for local languages are being developed, such as those being used in Ethiopia to reach different communities.

While the continent has made some progress, its share of the global AI market is still too small to actualise the $1.5 trillion market opportunity forecast by the United Nations’ Economic Commission for Africa over the next eight years. Failure to address this gap in AI adoption means that Africa risks treading a path of slower tech adoption, which has already proven to have detrimental and compound effects. So, what can progressive leaders in Africa do to bridge the adoption gap?

Government policy has a strong role to play in shaping and accelerating AI adoption. In this report, we analyse existing policy levers against a set of objectives that are most relevant to governments in Africa. Based on our analysis, we recommend that African governments:

1. Take a strategic adoption approach to identify, develop and test AI use cases that aligns with a country’s key national objectives, resources and capacities. While policy has an important role to play in AI adoption within a country, not all policy levers are suitable for all national contexts. The strategic adoption approach allows countries along a spectrum of AI readiness to embrace and pilot AI use cases responsibly and effectively, and within the current capabilities of the technology – not the hype surrounding it.

2. Prioritise responsible AI, defined globally as AI that promotes inclusive growth and is human-centred and transparent. The development and procurement of AI in Africa must be underpinned by these tenets to unlock the social and economic benefits of the technology for all.

To unlock AI adoption, governments need a set of tools and systems. While indices and guidelines already exist, there is not a comprehensive roadmap available to countries embarking on their AI journeys. At the Tony Blair Institute for Global Change, we are working with governments on AI-related issues, potential partnerships and a framework for delivery. Within this report, we provide a toolkit that guides policymakers through the questions, processes and activities that can be used to accelerate and shape AI adoption in the short to medium term for positive economic and social impacts over the long run.

AI is reshaping public services. It enables more efficient and effective approaches through process automation, faster customer responses, operational efficiency and predictive technologies. Such systems are already in use across developing countries to help detect counterfeit drugs, improve farmers’ decision-making and agriculture yields, and help deliver welfare benefits more equitably and efficiently. When combined with machine learning (ML) and data analytics, AI tools can help leaders derive new insights and improve decision-making.

According to recent research, AI is the basis of technological improvements that would enable meeting 134 targets – equivalent to 79 per cent – of the UN’s Sustainable Development Goals (SDGs). Achieving these SDGs in Africa could open up $12 trillion in market opportunities and create 380 million jobs by 2030 for the world’s youngest continent (by median age).

Other large-scale AI opportunities particularly relevant to Africa, but which could create significant benefits for the rest of the world, include:

• Decreasing the digital and information gap by using natural-language processing capabilities to translate and preserve some of the 2,000-plus languages in Africa – the world’s most linguistically diverse continent.

• Digitising, curating and sharing Africa’s rich ancient history by using ML, a process that has made the Timbuktu manuscripts widely available on Google’s Arts & Culture Hub, “Mali Magic”.

• Strengthening wildlife and conservation efforts in Africa, which is home to several biodiversity hotspots, through cloud-based databases and AI.

With such opportunities on the horizon, policymakers in Africa who want to deliver economic and social benefits to their citizens – and further afield – should consider not only accelerating but also shaping AI adoption in Africa. Adoption should not be limited to the private sector, however, with governments equally well placed to take advantage of the opportunities afforded by AI.

Despite these opportunities, many African nations and other least-developed countries in the Global South are not poised to take full advantage of AI, according to a PwC report titled “Sizing the Prize”. Africa, Oceania and some Asian markets (beyond the developed parts of the continent) are expected to benefit the least from AI, with the total impact of AI estimated at 5.6 per cent of GDP or $1.2 trillion by 2030 – compared with 26.1 per cent of GDP or $7 trillion in China and 14.5 per cent or $3.7 trillion in North America. According to the report, “developing countries will experience more modest increases due to the much lower rates of adoption of AI technologies expected”.

While the potential to accelerate AI adoption exists, emerging market data support PwC’s modest projections about the future of AI adoption in Africa. In the last quarter of 2021, global revenues from the AI software market were expected to reach $36 billion, with this figure projected to rise to $118.6 billion by 2025. The AI market in the Middle East and Africa is estimated to have been worth $870 million in 2021, with only marginal growth expected in the next three years.

Investment in Africa’s AI startups is also in its infancy. Although 2021 was a record-breaking year for tech investment – valued at $2 billion – across the continent as a whole, only 4.4 per cent of this total went to AI companies (one firm represented 89.9 per cent of this sector share) – compared with the 48.3 per cent that went to 184 fintech companies. Beyond market size and funding, Africa lags behind other continents in government AI readiness. Countries in Africa rank lowest on the Oxford Insights Government AI Readiness Index, a global metric that ranks countries by how prepared their governments are to use AI in public services.

Africa was left behind during phases of the Third Industrial Revolution, with the compound effects still being felt today. Africa cannot risk missing out again on the technological progress promised by the Fourth Industrial Revolution: the cost, especially to the continent’s young and entrepreneurial population, will be too great.

Creating an ecosystem that enables AI to thrive will deliver benefits to economies, society and governments, with emerging evidence showing how policy can both shape and accelerate a country’s AI adoption. Prominent indices show a clear correlation between national AI strategies – the policy document that governments use to convey their AI interests and ambitions – and the impact of AI on a country’s economic activity. Indeed, the Oxford Insights Government AI Readiness Index identifies a government vision, articulated by a national AI strategy, as a core component of a country’s AI readiness.

Stanford University’s Global AI Vibrancy Tool also demonstrates a similar link: countries scoring higher in AI vibrancy, or which have more robust AI-related research and development and economic activity, are those that have well-established national AI policies. Additionally, Tortoise’s Global AI Index will only rank countries in its index if they have specific AI policies, since government policy is considered a key component of related investment, innovation and implementation.

The World Bank’s review of national AI strategies and policies confirms the value of clear AI policies: “​countries that have seen early and broad adoption of AI such as South Korea, Canada and China demonstrate another common feature: a national strategic direction for AI defined by their governments”.

But AI policy not only accelerates AI adoption, it also shapes and informs its impact. Around the world, AI-related harms are becoming a major part of the policy debate among policymakers who are concerned about reducing risks. In AI-advanced countries, research shows that people of colour, children and other vulnerable groups face disproportionate risks from AI harms and, increasingly, policymakers are aligning on regulation to tackle them. Countries with coherent AI policies have an opportunity to lead the global conversation on how to mitigate these harms.

Over the past few years, we have seen the fallout from several failed and harmful AI-related government interventions, including the UK’s A-level grading scandal in 2020 and the Dutch tax office’s unregulated use of algorithms, which led to discriminatory and racial profiling in 2021. Incidents such as these have raised questions about responsible AI use by government, which policymakers have scrambled to address after a public outcry. These scandals do not exist in isolation: they undermine confidence in both the government and the technology itself. More effective future public programmes might be at risk because people simply don’t trust them.

What is Responsible AI?

There are many definitions of responsible, ethical or trustworthy AI, but the most influential international efforts have sought to define responsible AI through alignment with certain values and principles, such as those set out in the United Nations Educational, Scientific and Cultural Organisation (UNESCO) Recommendation on the Ethics of AI (2021) and the Organisation for Economic Co-operation and Development’s (OECD) Recommendation of the Council on Artificial Intelligence (2019). Leaders and policymakers seeking to build a strong AI framework that harmonises with international norms should adopt globally accepted responsible AI principles, adapting them when needed for the local context, rather than investing time in trying to build them from scratch.

At the least, the following OECD principles should be the foundation of a responsible strategic adoption framework:

• Inclusive growth, sustainable development and wellbeing

• Human-centred values and fairness

• Transparency and ease of explanation

• Robustness, security and safety

• Accountability

The UNESCO recommendation rightly takes these ideals a step further, articulating in more detail what these principles should look like in practice and expanding upon fairness to include equity and inclusivity.

Although the case for the economic value of responsible AI has not been covered in the same depth as the case for AI in general, it is crucial for unlocking economic and social dividends at a country level. Responsible AI will be the key for leaders and policymakers to enable transformative public-sector delivery through safe, secure, and trusted user and citizen engagement, securing foreign direct investment (FDI) through multinational partnerships, and assisting in talent attraction, retention and engagement.

Establishing Trust for Public-Sector Delivery

High-profile failures in the public use of AI tools risk eroding trust in institutions. For example, while the United Kingdom and the United States are leaders on government readiness, they both rank noticeably lower in terms of responsible use of AI, according to the Oxford Insights Index. Consequently, according to the TBI Globalism Study (2021), the two countries have some of the lowest acceptance rates of the use of AI in sectors that include health care, justice and welfare. Responsible AI is essential to establishing public trust and allowing governments to accelerate their tech delivery. When agreed-upon values and public trust in technology are not established at an early stage, its future is called into question. For example, genetically modified crops have seen significant pushback and restrictions due to a lack of public trust in the technology and those who championed it. Despite recent technical improvements in gene editing, this trust gap has persisted.

Attracting Investment and Talent

To accelerate AI adoption in developing countries, governments rely on global partnerships. However, there is a reluctance in the private sector to invest in countries where there is regulatory ambiguity. A recent study by the World Economic Forum on digital FDI shows that a regulatory framework is the third most important factor for companies considering such investment. Moreover, Stanford University's Artificial Intelligence Index Report 2022 indicates that smaller countries, including New Zealand, Ireland, Luxembourg and Sweden, which are driven by strategic adoption, outrank the superpowers with the highest growth in AI hiring. Highly skilled workers, such as AI developers and some members of Generation Z, tend to be driven by ethical labour markets so developing countries that want to attract talent and accelerate AI adoption should prioritise responsible AI and the accompanying regulatory frameworks.

The world’s first national AI strategy was published by Canada in 2017. Since then, governments have been moving quickly to keep up with the pace of AI development through a variety of policy levers. Beyond national strategies – which can make coordinated planning a challenge and have limited value for countries with different AI readiness levels – these policy levers include grants, regulations and standards to shape the way AI affects market structures and societal outcomes.

More advanced AI nations, such as the United States, have acted through most of the available means listed below.

While these policy options are in regular use and present legitimate options for African countries, many of the levers are suited to more mature digital and AI ecosystems underpinned by significant financial resources from government. For example, regulatory action can prove resource-intensive and challenging for developing countries to implement.

Not all levers are suitable for all countries and existing policy options are not always appropriate or tailored to the specific challenges of developing nations. For example, while national AI strategies have become a common benchmark for evaluating readiness, they require continuity of a government, broad buy-in and clear, actionable goals to be effective. They can also seem out of touch for nations lacking the more foundational elements of digital infrastructure, from connectivity to data policies.

Governments in low- and middle-income countries are often looking for policy options that bring immediate and equitable economic benefits, and which are suited to their political and administrative circumstances. Context matters, and a government with a nascent startup scene will naturally need a different emphasis than one with larger, more established industry players.

In Figure 7, we identify three existing prominent policy levers (publish a national AI strategy, build institutional capacity, develop government-private partnerships), adding a fourth option of “do nothing” and a fifth of “utilise a strategic adoption framework”. These are mapped against the three key objectives below that are crucial for developing countries:

1. Delivering for citizens through responsible AI

2. Participation of local actors, for example, startups

3. Economic impact

As shown above, the three existing standalone policy levers to accelerate AI adoption do not have a consistently positive impact on the three key objectives of policymakers in developing countries. Meanwhile, “do nothing” is not a viable option for progressive policymakers who care about harnessing the tech revolution to deliver for their citizens. But a new approach – “utilise a strategic adoption framework” – shows more consistent positive impact across the three key objectives. Below, we present and make a case for following this new approach.

Drawing on global best practices for responsible AI adoption, the TBI AI Policy Toolkit provides a starting point for the development of a country’s AI ecosystem. It integrates salient elements of the five key objectives (political feasibility, administrative feasibility, responsible adoption, local actors, economic impact) so that leaders and policymakers can:

• Balance the country’s needs with an understanding of the technologies currently available.

• Foster local and international collaboration.

• Provide a low barrier to testing AI adoption (feasible at both the ministerial and presidential levels).

Along with the AI Readiness Scorecard, Systems Map and Procurement Guide, the AI Strategic Adoption Framework is a starter toolkit for policymakers and advisors located in countries with nascent AI ecosystems. The toolkit considers AI applications in their proper societal and technical context while aligning the values and priorities of governments with the real opportunities afforded by the technology – not the hype. In addition, the TBI AI Strategic Adoption Framework translates the OECD’s principles for responsible AI into meaningful policymaking by following these guidelines:

1. Focusing on use cases that foster public trust and are highly feasible while having a significant impact.

2. Investing in research, development and mechanisms that support guiding principles.

3. Gathering global input and diverse data sets.

4. Articulating clear return on value to the public in exchange for data.

We invite relevant stakeholders and interested parties to email their questions and thoughts to the authors of the TBI AI Policy Toolkit presented in Figure 1.

Lead Image: Getty Images