Why do banks need to think like a fintech, how does co-creation with corporate clients work in practice, and how can AI help companies achieve a self-driving treasury organisation? Tony Wimmer and JB Commans from J.P. Morgan Payments, and Ariel Kuperminc and Jordan McFarland from Evoqua Water Technologies, share their insights.
Technology continues to provide new opportunities for treasury teams to streamline their processes and improve efficiency. Banks, meanwhile, are drawing inspiration from fintechs to accelerate innovation and roll out new solutions for their corporate clients. So what does this mean for treasury teams, and how can treasurers play a more active role in developing the solutions they need?
Thinking like a fintech
In today’s fast-moving world, banks cannot afford to stand still. As such, forward-thinking banks are increasingly taking steps to transform their product teams into in-house fintechs, explains Tony Wimmer, Head of Data & Analytics at J.P. Morgan Payments.
Given that payments are the lifeblood of the world’s economy, and given the role that banks play in ensuring money moves across the economy, Wimmer argues that banks “have a true homefield advantage” when it comes to leading innovation in payments. “At J.P. Morgan alone, we move over US$9trn each and every day,” he says. “There’s a lot of data in there, and data is what powers artificial intelligence and products.”
Wimmer notes that banks have unique knowledge, from client needs to local regulations. Likewise, they have data on “who pays whom, how, when, and how often” for their corporate clients – this “Payment Intelligence” as Wimmer describes “ is very hard for anybody else to replicate.” In addition, he points out that banks have “tremendous access to talent,” not least because of the investments they have made in technology and artificial intelligence.
In order to bring all this to life, says Wimmer, banks need to “think like a fintech” and follow the same principles that have made fintechs successful. In particular, banks need to:
- Operate a state-of-the-art data product platform with a micro services architecture that enable teams to deliver prototypes early, get client feedback, and make changes with every sprint.
- Focus on the highest priority products with a well-defined product vision – and commit to multi-year programmes to deliver against this vision.
- Be “obsessed” with clients – client needs are at the centre of a product and “design thinking” is an important part of a product team’s capabilities.
- Embrace adaptive and iterative development in collaboration with clients.
Power of collaboration
Where collaboration is concerned, there is much to be gained by designing solutions in tandem with clients, says JB Commans, Head of Treasury Data Products at J.P. Morgan Payments. He explains that the bank works with a panel of clients that are excited to participate as thought partners in the cocreation of products.
While every product is different, the co-creation process involves a number of steps: in the first instance, research is needed to pinpoint and prioritise the problems that need to be solved and identify the right client for the product development. J.P. Morgan then works closely with the client at every step in the design process to make sure the original problem is being addressed.
“We rely upon this iterative process to make sure we address the right problems,” says Commans. “That’s key to the co-designing process – we want to solve a practical problem, and co-design is how we ensure this happens.”
Cocreating in practice
A recent example of this approach was a cocreation project with Evoqua Water Technologies, a leading provider of water and wastewater treatment solutions. Building on the success of other cocreation projects with Evoqua, J.P. Morgan approached the company to propose working together on a new AI-powered cash flow forecasting solution.
“J.P. Morgan is the lead cash management bank for Evoqua, and has seen roughly 90-95% of all the cash transactions we do as a global company,” says Corporate Treasurer Ariel Kuperminc. “To have them come forward and demonstrate a commitment to work with us in transforming our data into actionable information in a consistent solution only emphasises the value of this relationship for both parties.”
As with most treasuries, Evoqua places considerable importance on its cash flow forecast, which is used as the basis of many cash management decisions. “As a treasurer, you wake up in the morning and wonder where your money is, where it’s going and where you need to place it,” comments Kuperminc. “So having a software tool that goes out weeks and months gives us comfort in the decisions we make for borrowing and repaying.”
When it came to collaborating on the new forecasting software tool, the initial discussions about Evoqua’s cash flow forecasting processes and needs were an essential first step in the process. “Sometimes you feel like you’re being interviewed in order to fit into someone’s process that has already been designed,” says Kuperminc. “In this particular conversation, there were questions the J.P. Morgan team returned to several times about how we think about things and how we use our forecast.”
In subsequent stages of the process, J.P. Morgan continued to take a thorough approach, seeking feedback on the tool and making sure it was fully aligned with Evoqua’s needs, adds Jordan McFarland, Treasury Analyst. “Ultimately any feedback I gave resulted in more tweaks, and then J.P. Morgan would ask me for more feedback on how that reacted,” he recalls. “It was truly a collaboration, with a lot of back-and-forth between the design team and our treasury team.”
Benefits of AI-powered forecasting
Once Evoqua’s team was comfortable with the layout and features of the forecasting software , the tool was implemented as part of the company’s standard process for forecasting. The resulting tool makes use of a variety of different forecasting models, running an AI algorithm that uses historical transaction data to choose the best fit model for the company’s particular transactions.
“On top of the advanced AI, we have the ability to investigate any specific spike in the company’s history. You can click on any day and find out what happened that day transaction by transaction. This provides the layer of context you need to fully explain the numbers you are presenting,” notes Commans. “And if that spike was a one-off event that is never going to repeat itself, you can choose to take it out of the equation so that the AI won’t factor it in.”
Likewise, treasury teams might know that a large flow is due in three months’ time, and add that one-off to the results. “It’s about creating an end-to-end intelligent workflow solution, where all these layers of human knowledge can be integrated with the AI,” Commans says. “At the press of a button, all this information flows through the process.”
Evoqua uses both top-down and bottom-up forecasting techniques at a company-wide level, as well as a ‘back of an envelope’ forecast. Compared to other methods, the AI model “is very much more informed, keeping a lot of history and being able to look at the patterns we’re unable to pick out,” says Kuperminc. “It’s less prone to human error – and a process that would otherwise take hours now takes just two minutes.”
The road to self-driving treasury
Turning to the future, Wimmer and Commans envisage a future in which companies can use AI to build a self-driving treasury organisation.
“There is a version of our future where Artificial Intelligence is fully trained to recommend, and then even accomplish tasks for treasurers,” explains Wimmer. “In this bionic world, the AI executes day-to-day, simple tasks. It finds patterns invisible to the human eye, bringing attention to critical issues – so humans can focus on what they do best: finding creative solutions to complex problems.”
The cash forecasting solution, for example, not only compares cash flows over time, but also uses AI to build an added layer of insights on top of the company’s data. “The next stage will be to have AI recommending actions – for example, you could programme it to optimise your working capital, or decide what currency to send a payment in, to optimize your foreign exchange fees,” says Commans.
The final step – “and the one we’re all very excited about” – is to create a feedback loop whereby treasurers can indicate that recommendations made by the AI are beneficial, Commans adds. “And when you’ve collected enough information about this, the AI can start doing things on your behalf, which is really what we’re driving for.”
Evoqua’s McFarland agrees that the idea of self-driving treasury is a welcome prospect, noting that some types of recommendations could bring considerable value to many of the company’s routine processes. “If we could get to a place that our banking infrastructure just knows that we need to move X amount from one account to another on Wednesdays, without us having to be reactive to certain events, that would make our lives even easier,” he concludes.
