‘AI has already reduced transaction analysis time by 90%’ 

At a time when financial institutions are calling for greater compliance, artificial intelligence seems to be an essential tool. While the cost of AI may be an obstacle, Matthieu Andrieu, head of corporate banking at Finastra, a company specialising in the development of financial software, tells Célia Cuordifede that he believes that there are financing solutions available. 

Conversation with

by

This article is sponsored by Afreximbank

African Business: How is AI being used to address compliance issues?

Matthieu Andrieu:It’s a technology that’s very well suited to compliance issues, because it can very quickly identify information from a huge mass of data, which is the first task of compliance. AI has already produced good results in banking institutions and in compliance systems. In particular, it has reduced the time taken to analyse transactions and/or documents by around 90%. 

How has AI helped compliance evolve from a state of ‘reaction’ to a state of ‘proaction’?

Artificial intelligence has traditionally been used to monitor payment transactions and document exchanges in real time. This is known as “reaction”. The new generation of AI, known as generative AI, popularised by ChatGPT, has been able not only to read data and identify it, but also to link it together and generate content, including images and documents. The next evolution of AI in the compliance system will undoubtedly be to make it capable of producing audit reports, which will explain why a person is compliant with bank rules or market legislation. 

What do you see as the limits of AI?

Because AI is so new, there is very little regulation of it at the moment. However, we have already identified a number of shortcomings. One of them is transparency in the decision-making process. Today, we are unable to understand why an AI-based system has taken one decision rather than another. 

AI uses millions of parameters that are not always known, and which depend on how it has been trained. If certain human behaviours are sometimes difficult to explain, we find the same difficulties with AI behaviour. Finally, I would also mention the challenge of responsibility. The more technology evolves, the more decisions it will make in place of humans. And we need to have a legal or physical person ultimately responsible, particularly in the event of a lawsuit. 

To what extent is AI used in compliance systems on the African continent?

AI is used in the compliance systems of banks in Nigeria and South Africa, which have substantial investment capacity and specialised technical teams. The cost of using AI can be an obstacle, which is why funds have been set up (including one initiative of Afreximbank) to give African banks access to the latest compliance technology.