Since the early days of the development of computers in the 20th century, scientists have strived to develop systems that are as intelligent as humans. Even popular sci-fi films like Star Wars have hinted at the concept of AI through their depiction of human-like robots. Although, starting in the 2000s, large tech companies built many AI based products, such as search engines and personalized recommendations for popular consumption, it was not until the release of ChatGPT—a generative AI tool—that AI truly entered the mainstream, thereby highlighting its potential to disrupt our everyday lives.
But here’s the thing, generative AI models such as ChatGPT represent only a very small portion of how artificial intelligence can be used to optimize the way we currently do things.
While it’s true that AI can help contribute significantly to Africa’s development, the notion that generative AI will be the magic solution is ill-informed. Africa faces unique challenges that necessitate customized AI solutions built and trained on African data—both generative and non-generative AI—to drive development and most importantly, address the problems that plague the continent today.
In this blogpost, we’ll be taking a closer look at the crucial pain points that AI could potentially address in the financial services industry in Africa, while also exploring the limitations of generative AI as one of the many possible artificial intelligence algorithms.
Generative AI: The Limitations
Just like the name implies, generative AI is an artificial intelligence technology that generates content, whether it’s textual, image-based, or even speech-based content. It has the capability to understand language and logic, enabling it to break down complex ideas into simpler ones. Furthermore, it functions as a skilled co-pilot, assisting in brainstorming and facilitating the process of getting started from the dreaded empty canvas or overcoming creative blocks.
However, despite its strengths, generative AI is a probabilistic model, which generates answers that are not guaranteed to be 100% accurate. While it can perform greatly at initiation phases of ideation, it does not have a full grasp of cause-and-effect relationships. Additionally, generative AI applications are unable to understand and address their inherent biases and limitations, which are shaped by the data on which they’ve been trained. And this is why generative AI should not be relied on as a dependable source for current and trending information. For instance, generative AI cannot be entrusted with providing real time insights on today’s news, offering mental health advice to patients, or even delivering judgements in legal matters. Its true value lies in enabling and stirring up creativity from a blank slate. Generative AI technology, in its essence, is most efficient in content generation and is—as at the time of writing this in August 2023—constrained within these boundaries.
The Real Challenges: Going Beyond the Hype
While the possibilities of generative AI are indeed captivating, many of the critical issues faced by businesses are more fundamental than we perceive. For instance, in the current financial landscape in Africa, there are a plethora of challenges that we have seen to limit the efficiency of enterprises within the ecosystem. These are challenges that are better served by other types of AI models (non-generative AI), and some of them include:
• Poorly defined user identities:
During KYC (Know Your Customer) processes, financial service providers require their customers to go through certain checks to ascertain the identity of the customer that wants to make use of the platform. This poses a challenge most of the time as due to factors like limited documentation, errors, and poor data presentation, which then results in lack of accuracy in user identities. AI models for Computer Vision, Information Extraction, and Natural Language Processing can improve many parts of this process.
• Identity takeover fraud:
Africa is not a stranger to the challenge of fraudulent financial activities where fraudsters takeover and impersonate the identity of genuine users, thereby leading to financial losses for individuals and businesses. In cases like this, real time fraud detection solutions trained on African transaction data would be extremely valuable in curbing the potential of such criminal attempts. Deep Learning based models, like those developed by Pastel, can help banks and fintechs uncover complex patterns that would not be detectable by the human eye and any rules-based patterns.
• Inefficient user onboarding:
Users of financial technology services usually crave seamless and quick access to financial products and services, and the moment onboarding processes begin to get cumbersome, it leads to the problem of drop-offs. To stay competitive, fintech solutions are constantly seeking ways to optimize their user onboarding processes for speed. Information Extraction and Computer Vision models have the potential to streamline these processes to enhance overall user experience.
• Credit scoring:
Let’s face it, traditional credit scoring processes are extremely cumbersome, inefficient, and often lack accuracy, thereby leading to a situation where a good number of people who genuinely require loans and qualify for credit are left underserved. Machine Learning based credit scoring, built using both linear and complex, non-linear models, can utilize a broader range of data points to assess creditworthiness of customers and reduce non-performing loans more effectively than any traditional processes.
• Face verification:
The existing computer vision models designed for face verification perform poorly with African faces. For instance, this is why Automated Teller Machines (ATMs) in African cities sometimes struggle to recognize the faces of genuine African users, as most of these machines were not primarily built with African faces in context. This loophole reveals the pressing need for advancements in this domain as well.
The Opportunity: Pastel Leading the Way
At Pastel, we firmly believe that addressing these challenges, many of which are unique to Africa, presents a tremendous opportunity to drive Africa's transformation from within. We are at the forefront, actively developing solutions that tackle these problems head-on. By leveraging a synergistic combination of generative AI and non-generative AI technologies, along with millions of data points from African users to build our solutions, we aim to revolutionize how enterprises in Africa manage their finances, mitigate risks, and unlock growth potential by providing them tailored AI solutions that suit the African context, ensuring relevance and efficacy.
While there is still a long way to go, we are excited about the progress we've made thus far. Our real-time transaction monitoring system leverages cutting-edge algorithms, analyzing transaction patterns to identify and flag suspicious activities, safeguarding businesses from potential financial losses. Additionally, our credit risk assessment tool helps enterprises evaluate the creditworthiness of partners, customers, or suppliers by analyzing extensive data points.
Africa's distinctive challenges demand customized solutions, and as an organization, we are committed to doing our part. However, changing the narrative and making AI more accessible and impactful in Africa necessitates a collaborative effort. We invite you to join us on this journey. Together, we can shape a future where AI-driven technologies foster growth, fuel financial innovation, and empower every African enterprise.
