The benefits of A.I. in Sub-Saharan Africa

The benefits of A.I. in Sub-Saharan Africa


Artificial Intelligence is not immediately associated with Africa, a continent where c.60% of the population doesn’t have access to the electric grid1. A.I. has the potential to impact positively communities in SSA. One example of that is the use of A.I. in Microfinance.


Traditionally, financial institution assesses the creditworthiness of a client by using a credit score. This credit score2 is the output of an algorithm using as inputs record of the borrower’s credit history from a number of sources (including banks and governments) or personal data like a contract of employment.

In SSA, this computation of the creditworthiness is not applicable for 2 reasons. First, the informal economy although declining still represents 38% of the GDP and 66% of the total employment3. Second, the banking penetration is very low4 (less than 30%) and banks can only offer loan to a small portion of their clients, those that they can appraise the creditworthiness by using the traditional sources mentioned before.

Africa has one of the fastest growing mobile penetration (75% of the population have a SIM) and the smartphone use is on the rise. On their smartphones, customers can use Mobile Financial Services (MFS) like mobile money to send, receive or store money with their device5. There is 100 million mobile money user and that number has grown 30% per annum between 2013 and 20166.

FinTech companies can take advantage of the non-traditional data that can be extract from the smartphones and create a new credit scoring model.


MyBucks is a German FinTech company operating in Sub-Saharan Africa. The company offers mobile financial services such as credit through a digital platform. According to the company, their success is due to their “proprietary credit decisioning and scoring technology and self-learning algorithms”.

MyBucks has developed a credit scoring technology platform7, Jessie, that helps them assess the creditworthiness of a client. The platform uses traditional data such as transactional data or employment verification when available. Its singularity is that it also uses alternative data and behavioral data. A.I. main role is to analyze the alternative data coming from different sources like cellphone usage, mobile money or social media.

As reported by Dr. Christiaan Van Der Walt8, Chief Technology Officer at MyBucks “The system assigns a unique credit score and determines a probability of default, which in turn drives a unique credit offering to the client, by adjusting the loan amount, term and interest rate.”

CEO and Founder of the company Dave van Niekerk gives more detail on how the machine learning algorithm works9 “This credit decision and scoring system works to continuously learn customer behavioral patterns from historical loans and is able to very accurately predict a customer’s probability of default for a particular product at any given time.”

Haraka is the star product of MyBucks. It is a real time nano-loan mobile banking app that has been developed in-house. The app can offer loans to clients that don’t have a credit score or that don’t even have a bank account. According to Dr.  Van Der Walt, real-time nano-loans is “a product that would not have been possible without AI and technology since loan amounts are extremely small, and a system where human intervention is required for credit scoring and loan disbursements would not have been financially feasible”.


Scholars Erik Brynjolfsson and Andrew McAfee10 explain that the breakthrough of A.I. is due to 3 factors:

  • Substantially more powerful hardware.
  • Significantly improved algorithms.
  • Increased amount of data.

As long as software engineers have access to calculation power at an affordable price, they will be able to create better algorithms that will expand the horizon of the data that can be analysed.

The company’s strategy for short and medium term is to take advantage of this virtuous circle and develop better, faster and cost-effective software that will be useful for their financial services activities. Some examples of their latest projects are11:

  • FinCloud, their loan management system that uses A.I. to manage credit risk, loan book portfolios, and customer service relationship.
  • Dexter, a software that uses A.I. to calculate a client’s fraud score by comparing that client’s online behavior with past fraudulent behavior of known fraudsters


MyBucks has found a unique way to address the credit scoring challenge in Sub-Saharan Africa. To improve their machine-learning algorithm they will have to keep providing training data. The company will have to expand their customer base to have access to more data and keep developing new algorithms that will take advantage of the learnings from the existing ones.


Some ethics question remains. MyBucks is using data from the cellphones and social network of their customers. Is it an invasion of their customer privacy?


(788 words)




1 More people than ever now have electricity in Africa, but 600 million are still in the dark

2 Credit Score, Wikipedia

3 Leandro Medina, Andrew Jonelis, and Mehmet Cangul, “The Informal Economy in Sub-Saharan Africa: Size and Determinants” IMF Working Paper

4 Banking in sub-Saharan Africa Interim Report on Digital Financial Inclusion, European Investment Bank

5 Mobile in Sub-Saharan Africa: Can world’s fastest-growing mobile region keep it up?, ZDnet

6 Mobile financial services in Africa: Winning the battle for the customer, McKinsey

7 Company presentation, MyBucks

8 How FinTech Firm MyBucks Plans To Offer Access And Financial Inclusion To Africa’s Unbanked, Mfonobong Nsehe

9  Meet The Man Championing FinTech In Africa – Dave Van Niekerk, Founder Of MyBuck, Mfonobong Nsehe

10 Brynjolfsson and A. McAfee. What’s driving the machine learning explosion? Harvard Business Review Digital Articles (July 18, 2017)

11 Company presentation, MyBucks








All Politics Is Local: Why Boston’s City Government is Making a Smart Investment in Machine Learning


Additive Manufacturing in Medical Device: Can Johnson & Johnson stay ahead of the curve?

Student comments on The benefits of A.I. in Sub-Saharan Africa

  1. It was very interesting to learn how machine learning could be applied to assess creditworthiness in developing countries. The question about whether MyBucks is invading privacy is an interesting one. My opinion is that this is a real risk but I am not sure how to resolve it.

  2. This is a very interesting way of getting around the traditional inputs for determining the riskiness of a loan recipient. As noted in the piece above, the traditional inputs don’t tend to be effective in this region. Specifically, the fact that this system uses customer behavioral patterns to discern whether someone may default is very thought provoking. They gave online activity as an example of behavior that would be used by the algorithm. I wouldn’t have expected someone’s browsing history to be a reliable indicator of whether they could be trusted with a loan. Of course there are certain activities (searching “How can I get away with not paying back a loan?” for example) that would give some insight into this directly, but for majority of online activity I hadn’t realized there would be a direct-enough/meaningful link. Though of course they are pulling this along with other inputs such as personal loan history so perhaps once all these entities are packaged together it tells a more thorough story. If people start to learn what inputs the company is using I wouldn’t be surprised if they became very cognizant of what information they are allowing it to get hold of/how they change their behavior in order to seem like as good a candidate as they can.

  3. This is very cool! In addition to some of the sources of alternative data you mentioned for credit scoring, there is also a lot of exciting work happening around psychometric credit scoring that is highly relevant for the SSA context. Psychometric measurements can make a lot of sense in data-poor environments, where traditional sources of data (or even alternative sources) are less readily available, and can often be used in combination with other data sources. The way it works is essentially that a loan applicant will take some kind of test that includes behavioral questions and the traditional types of loan questions (income, assets, etc. ). Based on the results of the test, but also how much time is spent on each question, what key strokes you make (e.g. how you use an income slider), and other seemingly irrelevant data, the algorithm generates a credit score that can be used for loan decisions. An important consideration here is that it does open up the possibility of applicants trying to game the system. For this reason, it’s often necessary to switch up the questions and/or scoring on a regular basis. EFL ( is a company that has successfully applied these types of credit scoring methods across Latin America, Africa, and Asia.

Leave a comment