AI Goes Mainstream

Rapid Growth of Transaction Banking and Treasury Use Cases

Banks have been using AI for many years, deploying basic chatbots for customer service, for example. But there are far more sophisticated areas where the technology is being applied today. Indeed, numerous use cases for this technology exist in the corporate-to-bank relationship space that can benefit both sides of the relationship.

The hype around AI technologies has leapt to a whole new level in the past year. Recent progress on chatbots, making them much more responsive, and the arrival of ChatGPT – a form of generative AI – have introduced the topic of AI to a larger audience and opened their eyes to its true potential.

Crucially, however, many AI technologies are not new, they have been used successfully by financial institutions for some time. In fact, banks have already launched several successful initiatives, notably in the transaction banking space, offering a number of use cases that leverage AI.

Gautier Mouzelard, Global Head of Industrialisation for Trade Finance, Acquiring and KYC, BNP Paribas, outlines: “Within the world of transaction banking – and therefore treasury –the large number of client interactions, high volume of data, repetitive transactions, and operational teams involved in the day-to-day process, bring us perfect opportunities to add value with AI technologies.”

Transforming transaction processing

One of the clearest use cases for AI, he explains, is to speed up and digitalise transaction processing in banks by reducing the extent to which humans need to be involved. “A good example is invoice processing, which traditionally requires a certain amount of effort from both corporates and their banks. Here, AI can be used to facilitate payment processing through workflow automation and applying image recognition to documents to extract data and reduce manual exceptions,” he notes.

Another interesting use case has seen AI deployed on documentary credits, whereby BNP Paribas is using optical character recognition (OCR) technology, complemented with other AI bricks such as natural language processing (NLP) and machine learning (ML). “It’s very powerful because it is applied to a paper-based business where the manual workload to extract unstructured data is cumbersome,” reveals Mouzelard.

In this instance, AI now securely extracts more than 80% of data almost automatically while reducing the repetitive work of operational teams having to remove data manually, reducing the risk of error. The system also learns by itself with ML capabilities progressively enhancing the extraction ratio due to the repetitive nature of the transactions.

“The result is a significant improvement from a client’s perspective, as the processing time of their transaction can be sped up. At the same time, AI also brings the transaction processing workflow up to the highest possible standards, particularly regarding data quality and structure,” adds Mouzelard.

Putting the client first

Another area where AI can enhance the corporate banking experience is customer service. As observed, many banks already leverage basic chatbots to have conversations of a limited nature with clients. “These chatbots can respond to simple client queries and carry out basic tasks, such as creating or cancelling standing orders or direct debits, and giving more information on payments reporting when requested by clients,” explains Mouzelard. But clients want more than that, and BNP Paribas, too, wants to offer a greater range of capabilities, especially as the level of data that AI interacts with increases.

“AI technologies will progressively evolve to help the chatbots to derive more accurate responses, in particular in understanding the nuances of client requests and managing more complex tasks that treasurers grapple with daily,” he comments. “In the same spirit, client onboarding could be much easier in the near future by taking advantage of AI to process the large number of documents required for the know your customer [KYC] process.”

Applying NLP technologies to client onboarding, for example, will help to analyse the documents and propose decisions to a bank’s due diligence officers. They can automatically cross-reference sensitive information such as ultimate beneficiary ownership or corporate shareholders hierarchies with external sources, for example.

“It will lead to a faster and more efficient onboarding process with a much better experience for both the banks and their clients,” enthuses Mouzelard. “Some proofs of concept are underway within BNP Paribas with promising results so far. And these could further enhance our already successful digital onboarding tool, Welcome.”

Boosting controls and security

Fraud management is another top priority for banks and corporates where AI can play a significant role by enhancing the effectiveness of the control framework. The first technologies developed with simple rule-based approaches identified only a limited number of suspicious transactions across areas such as payments, trade finance, supply chain management, and factoring. Moreover, these technologies generated many transaction alerts where no real fraudulent or financial security risks were associated. But AI is changing this picture.

“The addition of AI layers has helped to drastically reduce the number of irrelevant alerts or false positives, thereby helping to ensure that members of the operational team focus their expertise and analysis on the sensitive transactions raised by the detection tools,” reflects Mouzelard. “At BNP Paribas, for example, the use of AI to detect payments fraud has delivered excellent results, with more than 70% of fraudulent transactions stopped. This solution is now also complemented with ML capabilities that will further enhance the detection rate.”

Similarly, the fight against money laundering and terrorist financing has drastically evolved with the use of AI technologies within the bank, as Mouzelard highlights. “Ideally, to tackle AML, the bank needs to process and gather a mass of data – not just internal bank data and client data, but also external reliable market data – to find the patterns that indicate illegal activities,” he explains. “This is where the AI technologies we have developed internally, alongside third-party RegTech [regulatory technology] AI solutions, have brought strong added value to contextualise alert detection.”

In turn, this has delivered additional valuable information to the bank’s AML investigators while limiting the number of alerts. “In fact, the excellent results identified from using these tools in the trade finance business convinced the bank to progressively deploy those solutions on other transaction banking products, such as supply chain management,” enthuses Mouzelard.

Better serving customer needs

Just as AI can help bank investigators to make better decisions around KYC or AML, it can also help commercial teams to suggest value-adding solutions to treasurers. “There is a simple principle to consider when assessing a bank’s product offering and the value it can bring: if the bank knows its clients well, it can make suggestions that best suit them,” explains Mouzelard. “It’s a similar concept to shopping on an online marketplace and seeing ‘people who bought this also liked this’ prompts. That is exactly where AI can add value in the bank product space.”

Using datasets analysed from the client’s product portfolio, size, habits, and financial data, AI models can automatically make product suggestions to the commercial teams, he says. “This will usually involve transaction banking products that similar corporates are using but that this specific client has not yet adopted. It could, for example, be a suggestion to use trade finance products to cover risks in countries where the client is currently using an open account approach.”

In this use case, the product suggestions are automatically delivered to the commercial teams having been initiated by the AI solution. “This workflow is already in place in some countries within BNP Paribas, and some insightful results have been observed so far, so we are excited to see how it develops and where we can leverage it across the bank’s global footprint,” notes Mouzelard.

Cleaning and greening data

Elsewhere, there is a similar potential interest in harnessing AI for ESG purposes. Emerging regulatory requirements, like the European Sustainability Reporting Standards (ESRS) and EU Taxonomy, together with internal ESG policies such as those aimed at reaching net zero commitments, are adding significant pressure to banks’ data management around correctly identifying and labelling ESG-related data. This accelerates the need for the right technology to manage that ESG data.

“Increasingly, banks will have to appropriately and accurately classify green transactions according to the EU Taxonomy,” affirms Mouzelard. “Once again, the application of AI can ease this process thanks to the additional layer of automation, processing, and analytical capabilities of this technology. And since ESG commitments are part of BNP Paribas’ DNA we expect to deploy AI further in this space in the near future.”

Embracing the robot

Despite the busy pipeline for forthcoming developments, AI technologies are already a reality in the transaction banking business today. “At BNP Paribas we have clearly identified successful use cases and have various ongoing proofs of concept for additional enhancements.” This is good news for both the bank and its corporate clients.

“As we have already seen, AI brings value for BNP Paribas with huge data-processing capabilities applied at various steps of the transaction banking value chain, while clients benefit from improved security and an enhanced experience,” states Mouzelard. “Even if it will not directly replace our traditional technologies, AI clearly complements them.”

Of course, the path to an AI-enhanced future will not be entirely smooth. While the use cases are stacking up, the pace and scale of AI adoption will also depend on the current challenges raised by this technology.

“To succeed in the transaction banking space, AI requires high data quality as input, as well as large data servers. There are also significant costs associated with AI, not only in terms of the tech infrastructure, but also the human expertise and data experts required to create and managed the models,” notes Mouzelard.

“But those are the challenges we face with any emerging technology,” he concludes. “The key is knowing when and how to invest in order to make the most of the possibilities, for the bank and for our clients. At BNP Paribas, we recognise that AI is progressively becoming a core aspect of transaction banking, and we are innovating to make sure we are ready for it, and that the corporate treasurers we support can harness the power of AI in a safe and controlled way.”