Home IssueArtificial Intelligence 5 Q’s for Abake Adenle, CEO and Founder of Ajala

5 Q’s for Abake Adenle, CEO and Founder of Ajala

by Hodan Omaar
Abake Adenle

The Center for Data Innovation spoke to Abake Adenle, CEO and founder of ajala, a London-based startup that is developing AI-based speech technologies for low-resource African languages. Adenle discussed the opportunities voice automation tools present for economic and social development in Africa and the obstacles to operating across multiple countries with little or no data privacy regulations.

Hodan Omaar: What does the market for voice automation in low-resource languages look like and how do you think it will change over time?

Abake Adenle: This is the big question ajala has set out to address. Voice automation is one of the earliest forms of AI to have achieved widespread consumer use. Many speakers of highly resourced languages like English are familiar with early awkward automated telephone services and increasingly sophisticated conversational agents able to handle complex dialogue. While we have examples of what has worked in the United States, Europe, and even more recently in China, the role voice automation can play in African markets remains largely unexplored. 

There are opportunities for quick wins. Voice automation tools can deliver convenience and facilitate service access in markets where large portions of the population are illiterate and have limited access to the Internet. Consider partnerships with financial service companies such as retail banks, payments providers, and insurance companies for example. This is where I see voice automation playing a potentially unique and impactful role in African markets, not only as a tool businesses can use to achieve economic and operational gains, but more broadly as a channel that sits alongside websites and apps as a means of information and service access at scale. 

The market proposition in Africa is unique, and there is much to learn from voice solutions that have achieved traction in the United States, Europe, and China. However, I believe success at scale in Africa will require products that are highly tailored to individual localities and contexts.

Omaar: Ensuring speech recognition systems are accurate requires access to large amounts of data. How do you go about collecting the data you need and what do the data privacy and data security hurdles look like in the African context?

Adenle: Establishing data governance frameworks that respect individual data rights and privacy is core to how ajala operates. This can be challenging as we operate across multiple African countries, many of which have no data privacy regulations, or are in the early stages of ratifying and enacting legislation. 

Low-resourced languages have, by definition, limited stores of readily available data on which we can train speech recognition and speech synthesis models. We take the approach of collecting high-quality data in a highly targeted, transparent manner. We also rethink our approach to training models to embrace efficiency as much as possible. Many modern large-scale speech recognition models are trained on tens of thousands of hours of audio data. While the quality of speech recognition models has improved, the rates of improvement are not always commensurate with the increase in volume of data required for training these models. This suggests certain inefficiencies. 

Our aim is to collect data in a socially responsible manner: paying speech-sample providers, training data collectors to provide them with transferable skills, establishing internal standards when operating in countries that have no formal data privacy regulations, and explaining data usage and collecting only the data we need to avoid the pitfalls of large-scale, data-hungry applications. 

Omaar: While AI adoption in many African countries is still nascent, it is certainly evolving. From your perspective, what efforts should be at the forefront of policymakers’ minds to spur adoption for economic and social development in countries like Nigeria?

Adenle: For a country like Nigeria, the changes required are far-reaching and touch almost every aspect of civil society and public policy. Education is foundational and largely underfunded, leaving Nigeria vulnerable to the implications of an underserved young populace ill-equipped to interact with ever-evolving technologies that are increasingly important drivers of economic growth. To reap the benefits AI presents, Nigeria, and African countries more broadly, must invest in public education systems that equip all members of society with the tools to understand, benefit from, and contribute to digital economies that are simultaneously local and global. This mandate extends beyond educating engineers and programmers, to ensuring the average citizen possesses a basic level of AI and data literacy.

Data privacy and data security have become touchpoints in economies where AI has matured, revealing pitfalls to be avoided and guardrails that should be put in place to promote equity and safety in digital interactions. This is especially important where people face the consequences of automated or AI-driven decisions. African nations must prioritize policies that preserve the data rights of African citizens.

Finally, when introducing regulations to manage private sector activity concerning AI, and technology more broadly, African governments should employ a light-touch approach and communicate intentions transparently. Innovation will play a vital role in the evolution of African economies. We see the immense impact payments and fintech startups like PayStack and Flutterwave have had, and continue to have across Africa. We also see, in tandem, the evolving dynamic between regulators like the Central Bank of Nigeria and tech startups, and the resulting frictions. Safeguarding citizens’ rights is essential. However, when introducing regulatory policies, regulators should leave room for innovation and growth without excessive oversight and bureaucracy. Articulating AI and tech policy clearly and transparently gives African governments an opportunity to build relationships that will ultimately benefit government, the private sector, and the public.

Omaar: In what ways are the risks from AI systems in Africa the same as in the West and in what ways do they differ?

Adenle: In Africa, as in other places, the primary risks AI poses at a high level are preserving human agency, privacy, and transparency. When we consider Africa’s unique context, we see an additional demographic risk: a consequence of economies where a significant portion of the population is rural, and often economically or socially marginalized. 

Technologies that operate at scale imply some degree of automation. In turn, automation is often tailored to “obvious” or well-resourced scenarios where an average or common solution is often “good enough”. The plurality of contexts across Africa means the average that is “good enough” in more homogenous societies may end up further marginalizing substantial portions of African communities if technology does not embrace localization in a meaningful way. 

The success payments companies have achieved is a clear example of the value-add of localization in an African context. Fintechs have attained substantial valuations on the back of payments solutions localized to address myriad regulatory and technical constraints. For AI to achieve similar value and product-market fit at scale in Africa, localization must be a priority.

Omaar: We’ve been talking about AI in Africa as if Africa is a single market like the EU, which is not the case (though it seems to be moving in that direction). To what extent do you think it makes sense to focus on AI policy at the level of individual nations as opposed to the continent?

Adenle: The EU presents examples of how a regional bloc can successfully advocate for the digital rights of citizens of member states. The bloc has shown how regulations can be deployed to the benefit of consumers in nations that may not have been able to achieve similar results individually. 

Theoretically, regional trading blocs like the Economic Community of West African States (ECOWAS), the East African Community (EAC), or more expansively the African Continental Free Trade Area (AfCFTA), can leverage their combined position to advocate for similar benefits for citizens of member states. However, as you mentioned, the AfCFTA is still in its infancy, and these institutions seem to be more concerned with advancing regional trade at present. For now, the foundations of AI policy are being established at the national level where countries like Ghana, Kenya, Rwanda, and Nigeria are introducing policies to address data privacy, data security, and initiatives that promote engagement with AI across the private and public sectors. Considering these early policies may influence future regional perspectives, policymakers should, as much as possible, prepare for a future that embraces a coherent and unified African position on AI.

You may also like

Show Buttons
Hide Buttons