NextBillion - An Initiative of the William Davidson Institute at the University of Michigan

The impact of artificial intelligence (AI) on employment in emerging markets is a topic of much speculation. Though many analysts have concerns about its potential to make some professions obsolete, others point out that AI and other tech jobs that support remote work can also be a source of new, stable jobs for many in the Global South — especially in rapidly growing countries in Africa, home to many of the world’s fastest-growing economies. For example, new jobs are likely to be created in the AI supply chain, as data annotators and other data service roles will be needed to shape the datasets that AI models require to operate effectively. With how quickly AI is developing, these jobs may look like a great opportunity: The global AI market is expected to grow to almost $830 billion by 2030.

However, not all AI jobs are created equal. If we want to build a future where AI lives up to its full potential, we must embrace AI practices that are both responsible and ethical. While these two concepts are similar, responsible AI evaluates a model based on its entire lifecycle, including whether workers involved in its development are fairly paid, while ethical AI is focused on a model’s alignment with specific values, such as whether or not it produces unbiased content. When developed responsibly and ethically, an AI model will be based on the work of fairly paid workers with a broad set of perspectives, which can help mitigate the problem of unwanted biases in the data, and refined by testing to ensure that the model is behaving according to social and legal expectations.

To create this future, companies and governments need to work together to build ecosystems that can support not just AI jobs, but AI employment — the kind of employment that provides pathways to progress and real economic growth. 

The Advantages and Disadvantages of Different Types of AI Jobs

Data annotation is the work done by humans to categorize and label data to make it easier for computers to understand, and it’s a key part of training an AI model — and a key source of AI jobs. There are two main approaches to AI data annotation: crowdsourcing it or hiring dedicated annotation workers. 

Working for a crowdsourced data annotation platform may seem like a good starting point to build the necessary skills for a career in AI. However, the drawbacks of the job can sometimes outweigh the benefits. For example, this kind of job is often “piecework,” in which workers are paid a fixed rate for each task, regardless of the time it requires. For a data annotator, this work can be as simple as labeling photos — but it can also involve more complicated tasks. 

For instance, LiDAR annotation, a way of translating 2D images into 3D data, is particularly critical in the development of autonomous vehicles, as it allows the AI model that’s driving the car to know how close it is to other vehicles or objects. These tasks can take up a lot of time, and they may not have the same payouts as completing multiple, less-complicated tasks in the same amount of time. Adding to this challenge, crowdsourced data annotation work frequently lacks formal training — or it may offer training that is confusing, or that doesn’t fully prepare workers for the tasks they’re expected to perform. 

At its core, crowdsourced annotation work is contractor work, with all the disadvantages that often entails. Just like gig workers in the U.S., data annotators working for these platforms don’t get benefits and (as mentioned above) they are paid per completed task. But though these workers may get used to earning a certain amount per task, coming to depend on that consistency, pricing for these tasks can be extremely volatile — depending on the number of workers and tasks available, as well as the typical rates clients who contract these platforms pay for these projects as a whole. And these contracts are often made through a “reverse auction” process, in which there is one client buying, and many platforms are trying to lower their pricing as much as possible to secure the contract. So one platform may undercut another, directly affecting how much they receive from the client — and how much they pay out to the workers on their platform.  

Moreover, these workers are often remote, so they have to pay their own expenses when it comes to equipment and internet access — which can quickly add up. Yet if they decide to save money by using public internet, which can be easily breached, for their work tasks, this can represent a security risk: The data these workers are annotating can be extremely sensitive and valuable to the clients who are paying the platform for these annotations, and an AI model requires high-quality data to perform to expectations. And naturally, if the platform decides to shut down suddenly, these workers are out of luck — comparable to what happens when Uber or Lyft exit a market, causing their workers’ income to disappear from one day to the next. 

The Value of Stable Formal Employment in Africa’s Tech Sector

Financial instability can be a major factor in a person’s ability to lift themselves out of poverty, in part because it divides a person’s attention. To reduce poverty, governments need to invest in the infrastructure, policies and regulations that can support the growth of stable work, especially since most employment in Africa (about 86%) is informal. 

The data annotation work we discussed above is only one part of the AI supply chain, but it’s both time-consuming and critical. Up to 80% of a data scientist’s time is spent merely managing data, not analyzing it, and Sama has found similar ratios in our work with clients who contract us to annotate the datasets used to train their AI models. It’s estimated that up to 87% of these kinds of data science projects never make it into production, and a lack of sufficiently representative data is a major factor in their failure. Even if an AI model does get deployed, flaws in its training data may cause it to behave incorrectly. A self-driving car, for instance, may not be able to confidently recognize a motorcycle if it hasn’t learned from well-annotated data. In other words, there will always be a need for data annotation, and there will always be a need for similar digitally focused jobs — and this need could help generate the kind of stable, formal employment that can reduce poverty and drive lasting economic growth. 

A growing number of countries are already capitalizing on these AI job opportunities as part of their ongoing digital transformations, many of them through long-term plans. For example, Madagascar’s Digital Strategic Plan has already resulted in its digital sector generating 2% of its national GDP as of late 2023, totaling €365 million, and that number could be as high as 6% by 2028. Uganda, one of the countries where Sama operates, released its own Digital Transformation Roadmap last year and noted that Information and Communications Technology (ICT) already makes up 9% of its GDP. And other East African countries, like Ethiopia and Mozambique, are even collaborating with international entities in the pursuit of their digital transformation: The EDISON Lighthouse Countries network and the World Bank, respectively, have provided input and support for these countries’ plans. 

Clearly, these countries all have a desire to work towards a brighter digital future. How then do we as companies act as good partners? 

Kenyan president William Ruto has perhaps the clearest explanation of African countries’ current response to that question: “We as Africa have come to the world, not to ask for alms, charity or handouts, but to work with the rest of the global community.” The governments of Africa are asking countries to move beyond the “patronizing” idea that the continent needs aid. They want to present their countries as ready and willing to meaningfully participate in the digital global economy. This may seem like an insurmountable task, but companies can play a huge role in advancing this paradigm shift. 

How Tech Companies Can Foster Development in Local Communities

At Sama, our work focuses on AI model evaluation and data labeling; for many of our employees in our offices in Kenya and Uganda, working at Sama is their first formal job with a living wage. We have learned from our work with these countries’ governments that companies that take the time to develop local relationships can build trust that wasn’t previously there. We are seen as partners integrated into the local ecosystem, not massive, faraway companies with no understanding of the local culture or circumstances. In short, these relationships help foster a new era of collaboration that leads to better outcomes. For instance, Kenya’s government is setting up more ICT centers across the country based on our positive working relationship. 

It’s easy to talk about collaborating with local communities, but what has this looked like in practice for Sama? The first piece of our efforts to develop these local connections is hiring employees, rather than depending on crowdsourced, contract labor. By creating long-term and steady employment opportunities, we can offer a pathway into the formal economy — and into financial stability. Hiring also allows a company to establish roots, positioning itself as a reliable partner committed to growth. With countries like Uganda working towards ambitious development goals that heavily focus on ICT and digital transformation, that sort of thing matters. 

The second piece of these efforts, which is equally important, is abiding by both local structures and laws as well as global labor standards — an area where the AI industry is unfortunately falling short, particularly in emerging economies. For example, contractor work frequently doesn’t require an employer to pay payroll taxes, which can place an additional financial strain on people. Settling in, paying the required taxes, and following the employment laws of the country you’re operating in are all important ways a company can provide the reliability that governments are looking for in their corporate partners. Since those taxes also then get reinvested into the country and its infrastructure — resulting in better roads, more internet connectivity and more funding for education — they can also contribute to more growth opportunities for businesses themselves. 

These efforts to support local communities may sound simple, but they are not easy. The AI sector in Africa is currently wrestling with a “race to the bottom” mentality, where vendors are willing to undercut their competitors to secure a deal. Therefore, these businesses can’t or won’t pass the fruits of their success along to the employees who are doing the work for them, as this would reduce the competitiveness of their pricing. This obviously impacts workers’ wages and contributes to systemic poverty, but the effects reach further than that. Investing into countries’ infrastructure by paying taxes, renting office space, and making other efforts to join the local community helps create jobs at more than just one company. When taxes fuel a new road construction project, workers are needed to build the roads. Those roads then connect more people to employment opportunities and allow businesses to serve more parts of the country. Those growing companies will also need employees, and the virtuous cycle continues from there. 

Thinking Beyond the Bottom Line

Building a sustainable ecosystem in East Africa — not just for AI but for any technology — requires companies to put their money where their mouth is. For instance, big tech companies that decide to hire local workers can’t champion their ethical supply chain or boast about their corporate social responsibility on the one hand, then base their contracting decisions on a reverse auction process as described above. A reverse auction disadvantages companies which have formal employees; they can never be the lowest bidder due to the fact that they pay stable wages. 

It’s a struggle for companies to strike the right balance between doing the right thing and doing the profitable thing. It requires them to think beyond their bottom line and prepare to play the long game in navigating increasingly long-term considerations like social and environmental responsibility. But when it works, the results are beautiful. At Sama alone, we’ve helped lift over 65,000 people above the poverty line in East Africa. Imagine what larger corporations operating in the region could accomplish if they took a similar approach. 

With how rapidly AI is developing, the growth projections shared at the start of this article may be out of date sooner than we think. AI could be worth much, much more than the already-massive $830 billion it’s projected to reach by the end of this decade. And that doesn’t even account for other technologies beyond AI that are currently in development — or that haven’t been invented yet. 

Considering the needs of workers and communities, and taking the time to work together with local governments — while shouldering the responsibility that comes with these partnerships — can allow these emerging innovators to produce better results for their businesses, for the broader AI sector, and most importantly, for the people of East Africa.

Wendy Gonzalez is the CEO of Sama.