When it comes to artificial intelligence, banks are at a turning point. They are moving beyond the mere use of customer service bots and stepping into a promising new environment. Advanced AI applications are helping them do everything from predicting the market to stopping fraud.
Although the opportunity is vast, banks face resource- and data-related challenges in implementing such tech. Getting it right is crucial: The banks that excel with AI will gain a massive competitive edge over their industry peers.
To get a sense of where the industry stands and how banks can capture the promise and real business value of AI, we talked to DDN’s Paul Wallace, for his perspectives on AI and the financial services industry. Here’s what he had to say.
Accenture wrote that banking is approaching a paradigm shift that could allow for banks to reimagine their businesses for a world in which banking revenues are increasingly decoupled from workforce head count. Do you agree with that sentiment?
Paul Wallace: At a high level, I would agree that there is a new set of capabilities which, when properly used, will definitively help banks—including retail, finance, investment, and currency exchanges.
Looking more closely, banks are increasingly about business-to-business, machine-to-machine transactions, and the number of transactions they’re seeing is scaling up at the same time the market is becoming more competitive. As transactions get shorter, margins get even smaller. So, there’s this interesting playoff between the cost and the advantage of computational assistance, because the number of transactions is going up so fast—you’ve got micropayments, bank-on-bank transactions, investment transactions even within a fund. Generally speaking, the faster financial services organizations can make investing and banking decisions, the greater their returns.
Right now, we’re seeing that without AI-assisted decision-making, banks will be less competitive. AI allows these banks to not only make smarter and faster decisions but also more efficiently delegate tasks. The technology can raise an alert when a human employee needs to step in or review a transaction. Meanwhile, it automates tasks where humans aren’t necessary, thus driving down workforce costs.
Banking is becoming an increasingly fractured market, with competition from both traditional players and new, digital-first companies. Given this backdrop, how should banks be thinking about AI to position themselves for success?
Paul Wallace: First off, they should know that it’s vital—you aren’t going to succeed in this competitive market without the help of automation. AI is already creating a massive advantage for the banks doing it right, and that advantage will only become greater in the future. But it’s also not necessarily vital that you do all things at once. Artificial intelligence adoption requires a series of well-thought-out steps, and sometimes banks go wrong when they try to dive in without setting a strong foundation. So, prepare, do what you can, and know that as you get deeper into the process, AI becomes both easier to institute and more valuable—assuming you’re going about it in the right way.
What are some specific AI use cases that provide return on investment for banks, and where is this headed in the future?
Paul Wallace: For some years, we’ve seen automated customer service that finds information you ask for. That sort of interaction is possible because the data is open-source—it’s the language we speak. Anyone can train natural language. So, banks that want to interpret large amounts of contract data or documents, such as asset positions or company reports, can use pretrained models to read all that data, gather information, and present outputs in an interactive way.
But banks are now finding incredible value in a variety of other use cases, while building toward other areas where AI will have even more impact in the future. AI can have a really high ROI in risk management, price prediction, and fraud detection.
Let’s start with fraud. You’ve got retail fraud based on individuals and business entities—which entities are low-risk and which ones are fraudulently representing themselves as such. But we also see market fraud in price manipulation—where stock pricing can be influenced fraudulently, or wash trading to inflate the perceived demand. AI can identify behaviors and flag transactions that are potentially fraudulent.
Another interesting use case for AI is price prediction. People have always tried to analyze the behavior of ticker prices. But now, they are using AI language models to find long-term patterns that arise in pricing—the idea being they can now start to predict pricing 30 seconds or a minute in advance with some precision. That micro-pricing allows you to do high-speed trading and get a few points extra margin every time. It doesn’t sound like a lot, but another couple of margin points is fantastic at high-frequency trading.
And finally, when it comes to risk management, AI can help banks assess all the trades they’ve made during the day, work out what their exposure is to certain types of security or stock or currency based on those trades and their overall portfolio, and then make some intelligent decisions about where to place overnight reserves to minimize potential loss.
What challenges do banks face in putting some of these solutions in play?
Paul Wallace: Data to train the models. You can’t train a model unless you’ve got the right data, and in our recent DDN user group, one of our speakers said, “Data is the source code for AI.” One advantage that most traditional banks have is that they’ve been around for years, and they have all the data. An emerging competitor doesn’t have the history, the background, the information—they can use only public source data.
So, AI is a tool for everyone, but for some of the revenue-generating use cases—such as price prediction or risk management—existing banks have the relationship codified in the data already that they can now start to learn from and apply. This idea of backtesting—which investment banks use to rerun historical data against their AI model to see whether the AI would produce a better result—works much better when you have a longer history.
Another challenge for emerging banks is that AI costs a ton of money to do properly. If they don’t have the time, the runway, or the money to invest in the equipment or the renting of the equipment on the cloud to build and run these models, they’re not going to be able to do it. It takes a lot up front to build some of these AI systems. Traditional banks can afford that sort of investment, but they’ll want to ensure they’ve done the requisite preliminary work in securing their data foundation before making that investment.
Separately, there’s also a skills challenge: identifying people to hire or redirecting existing talent to help build that model and tune it for your use. It’s another significant investment, which includes all the time it takes to put all the equipment in place to get your AI application up and running. And some of those things mean that, if you’re an emerging or startup bank, you are going to have difficulty building those systems to run yourself.
So, considering these challenges, how can financial service organizations achieve success with AI?
Paul Wallace: First and foremost, DDN brings expertise that can guide companies in preparing for implementation. Successful enterprises use a data-first strategy, thinking through where their data is going to come from on the front end. They’re also very clear about the business outcomes they’re seeking. So, we help companies think through those outcomes and lay a foundation to achieve them.
So many AI projects don’t make it to production. This is due to a variety of reasons, not all of them having to do with the technology involved. Often, it’s about governance, executive leadership, or expertise. We can help with the entire spectrum. We work with a series of systems, partners, and software optimized for the business of AI, and we’ve encapsulated those recommendations in our AI Success Guide.
When you start dealing with a petabyte of data with hundreds of billions of parameters, you need to think differently about how you’re going to manage that data. You need to think differently about how you’re going to process the data. And ultimately, you have to figure out how to institute AI that helps you make business decisions that create real value for your organization. It’s the difference between having some solar panels on my roof and having a big solar farm down the road. We’re going to improve time to market, eliminate prototype risk, and secure scale-up.
But, again, if you don’t have your business objectives in mind, you won’t create anything useful. So, it’s important to have expert data scientists and engineers, which is what we can provide at DDN. That’s how you get the best outcomes and drive the greatest ROI.
Interested in learning more about AI in finance? Discover more insights here.
