How AI and payments data will transform corporate banking
Walid Koleilat Walid Koleilat
December 12 8 min

How AI and payments data will transform corporate banking

This article was co-written by Walid Koleilat, Andrew Marble and Cory Salveson.

The world of banking is no stranger to AI. Retail banks already use it to glean insights into customer behavior using data such as payment transactions. By better understanding important events in their customers’ lives, such as house purchases, banks can provide more suitable advice and better tailor their products. One fintech even claims it can predict customer divorces or relationship breakdowns.

With AI-driven transformations accelerating across all areas of the banking sector, corporate banks are following suit. In fact, the explosion of payments data combined with advancements in AI techniques is opening the door for a new wave of opportunities in corporate banking. Adoption is defining organizations’ competitive edge and even their viability in a rapidly changing world. The question isn’t whether AI in corporate banking is needed, but when to adopt it and how. From our work with corporate banking clients and discussions with industry colleagues, we’ve come to believe that the time is now. Corporate banks just need help with the “how.”

It’s still early days. AI is gaining traction for anti-fraud and anti-money laundering, but nobody knows all the ways AI will be implemented in corporate banking yet. In this article, we’ll outline a promising approach that uses current technical capabilities to realize an emerging industry trend: commercialization of corporate payments data.

First, it’s important to understand the “why” of AI in corporate banking.

An increasingly data-rich world

Banks are data-rich institutions, but traditionally, the interactions between banks and their corporate customers have been based on the relationship manager’s experience and knowledge of their customer’s industry—not on high-quality data. This expertise relies on extensive manual research and ad-hoc analysis.

But finance is becoming more complex than ever. Data-rich payments are occurring so fast that they’re almost in real time, speeding up trade flows. This data explosion is fueled by the frictionless exchange of goods and the increasingly elaborate web of connections across economies and previously distant trade partners. In consumer banking, new startups like Oak North in the UK, or Square Capital in the U.S., are using new sources of payments data to underwrite loans for an underserved market niche: small businesses that are profitable, but lack the multi-year track record to earn the trust of traditional banks.

This new class of customer insights solutions, built on new data flows, is advancing the industry—but what about relationship managers in corporate banking? They’re losing the ability to sift the signal from the noise, process relevant information and extract meaningful insights to serve their customers. There’s simply too much data for them to deal with.

With the adoption of ISO 20022, corporate payments will soon become even more data-rich. It’s a highly necessary standards shift that will lead to more details of each transaction being included along with existing payment data. This creates even more opportunity for banks to gain greater insight—or lose the forest for the trees.

That’s where the power of AI comes in. AI excels at taming complexity, so it can play a fundamental role in making sense of this deluge of data. However, implementing AI in corporate banking is more complex than in the consumer realm.

The value from the payments chain

The impact of AI is easy to see for consumer banks that serve individuals and small businesses, but harder to achieve for corporates.

As consumers, we provide data when we use banking services, and in exchange, these services become increasingly personalized and convenient. This balance is possible because although consumer spending data is increasing, the complexity of using it to tailor the service to consumers is relatively straightforward. More data just means that the bank knows more about the customer. The new payments data captured by Square, for example, is foundational to Square Capital — a separate venture founded in 2014 that has underwritten over $1 billion USD for small businesses. And that’s just one use case on top of newfound data.

In contrast to consumers, a corporate customer’s behaviors and needs are defined in the context of a supply chain involving multiple entities, such as producers, vendors, warehouses, transportation companies, distribution centers and retailers. Therefore, to deepen our understanding of corporate customers, not only do we need more data about the corporation itself, but also more data about the entities in the supply chain and the underlying payments connecting them. Getting all this data is challenging since the data is naturally stored in disparate systems, with records unevenly reconciled or connected.

In the big data and analytics wave of the last few years, banks have tried combating this complexity by funding large-scale data lakes, but they’ve not seen a return on investment to match the hype. This is precisely where AI can start to make a difference.

How can AI help corporate banks? The case of payments data

The outcome you achieve from AI is always dependent on how you adapt the technology to your business, but also how you adapt your business to AI. For relationship managers working at corporate banks, this firstly means finding techniques that can help them do their job better in an interconnected and data-rich world. Next, it’s a matter of designing the right ways to bring these techniques into your workflow to achieve the highest impact.

One emerging concept in AI that’s becoming part of the state of the art is representation learning, sometimes called feature learning. Using this technique, AI practitioners can use flows of payments data to create a data representation of corporate customers that acts as a set of key attributes characterizing how they are behaving relative to each other and in the context of their supply chain. This characterization can power analogical reasoning and be a source of insights for decision making.

While traditional approaches to data analytics rely on manually-defined features of a corporation, like industry or size, representation learning discovers attributes of corporates that are similar or different in the ways they are transacting. The technique relies in part on leveraging data from a large group of companies together so that similarities can then be used to inform where a new proposition (e.g., product, service, advice) would work well.

In our work helping corporate banks on their AI journey, creating representations using payment transaction data has been an integral technique for capturing valuable patterns and insights efficiently. Relationship managers already build mental representations of past and predicted future financial behavior; AI makes these representations richer and faster. Using AI, relationship managers can then achieve a deeper understanding of their corporate and institutional customers. Over time, these insights can improve the level of service beyond anything previously possible using financial statements alone.

This example of using AI and payments data to help relationship managers demonstrates that it’s possible for AI to help transform corporate banking, one relationship at a time. But there’s a big difference between the possible and the real. How can banks get started with opportunities like this?

Making it a reality

We frequently advise clients on how to make their organization ready to harness AI. In these discussions, we hone in on five key dimensions where the organization’s energy is best spent. These are strategy, data, technology, people, and governance.

In a banking context, this framework is flexible enough to meet the banking industry where it’s at: with lots of data, but with little in the way of accessible, high-quality data or modern system integrations needed for machine learning development.

In our experience, the right starting point is to organize around strategic use cases that can be fueled with relevant data flows to show value quickly. With the right proof of value, support for changes in infrastructure or even business model can then be unlocked.

From there, banks can develop the building blocks for delivering insights and recommendations to more applications and business lines—so relationship managers can streamline business decisions and help tailor the relationship between banks and their corporate customers. Payments data is a great place to start because of its centrality to that relationship and because of the increasing availability of this data.

Representation learning is just one example of how AI can help unlock new levels of insight from payments data. There are many more opportunities already and more will emerge tomorrow.

Unlocking this potential isn’t just key to how successfully relationships are managed today, but a vital determinant in shaping the future of corporate banking as a whole.