ChatGPT, and to a lesser extent other Large Language Model (LLM) AIs, has led to explosion of interest in AI. These exciting developments demonstrate the still evolving potential of ‘chat-bots’ and other AI that can mimic and replace human interactions. These innovations create huge opportunity, as well as some risk, for banks and other financial institutions. The utilisation of AI to automate, streamline and enhance many services, and so reduce costs whilst improving customer experience, will be a key competitive advantage for the bank of the future.
But AI and advanced data analytics are already helping many banks by automating middle and back-office processes. Fraud detection is one area where AI has already had an impact, and where new innovations can deliver significant new value.
BANKS’ DATA ADVANTAGE
When it comes to AI, banks have a crucial advantage over many other sectors. They already have masses of data. Data is the crucial fuel for AI. Billions of rows of anonymised transaction data, linked to channel, device, time of day, location and many other data points are needed to train new models. Banks have all of this. The challenge is creating the data platforms and consistent data models that allow AIs to use it.
Ensuring the quality of that data is also vital in creating effective AI’s that make good decisions. As Martha Bennet, a principle analyst at Forrester told TechTarget earlier this year "One of the things AI needs is lots of data, and banks have lots of data." She goes on to say in the same interview that; “Companies with well-structured, good data have already been able to put AI to good use in detecting fraud.” And that’s the secret. Creating unified data platforms that can operate at speed and scale with diverse, fast-moving data that can be relied upon as accurate.
BEYOND STATIC RULE-BASED APPROACHES
Banks have used analytics for Anti-Money Laundering (AML) and anti-fraud transaction monitoring for some time. But these have been static, rule-based systems: if this happens, then that, flag as potential fraud. These systems create a high percentage of false positives which add unnecessary friction into processes, increasing costs and damaging customer experience. They are also relatively slow and require frequent recoding as new fraud types and attack vectors are discovered. With the frightening increase in financial fraud and seemingly endless innovation in how they are perpetrated, enhancing fraud prevention with AI is an imperative.
The big difference with AI-based systems is not only are they faster and more accurate than humans or static-coded approaches, but they learn. They can spot new fraud patterns that have not previously been identified. Using deep learning techniques, they can identify data anomalies which would not trigger traditional fraud detection warnings. And more importantly, because they can monitor every interaction of every customer, all the time, they can flag behaviours which are unusual for a specific customer. Not only does this increase levels of protection, but significantly reduces false positives as every customer is different and treated as such.
Payments innovator PayPal was one of the pioneers of this approach. Using AI to run over 1,000 checks on data such as device ID, geolocation and repeat merchant purchases every time a customer uses the service to pay, it can instantly validate customer identity. What’s more, new models fed by new streams of diverse data, can be deployed in real time at any time.
FRAUD AND CX – TWO SIDES OF THE AI COIN
The connections between applying AI to anti-fraud and to improving customer experience are obvious. Better understanding of every individual customer and the more data it can learn from, the more accurate any AI built on understanding that behaviour will be. The quality of insights, recommendations and interventions will increase, and continue to increase with every interaction, whether that’s to suggest or prevent an action. So, customer experience and fraud prevention are in many ways two sides of the same AI coin.
Which underlines the importance of an integrated approach to rolling out AI across financial services. The current status of AI in financial services sees the adoption of a multitude of single purpose solutions, often implemented on a single department, product or geographic basis. Each leverages its own discrete data set for training and deployment. Different solutions, vendors and customers create silos of data that each provide AI with a different lens through which to learn. Specific AIs will be needed for different scenarios at least for the foreseeable future as so called General Artificial Intelligence is still some way off. But it is important, even at this stage, that AI applications access a common, consistent and comprehensive data set that represents as full a view as possible of an individual customer.
END-TO-END CRIME INTERDICTION
And this is the strength of Teradata. Working with banks to create powerful artificial intelligences that learn from and operate with multi-dimensional, scalable data in real time across the whole enterprise, we can provide the foundations for exceptional AI-driven fraud prevention, customer experience and a host of other applications. Looking towards the bank of the future we see AI as a core part of its operations, using a consistent, enterprise-wide data platform as their foundation. We are working with leading banks as they look to build on their current fraud detection solutions to create end to end financial crime interdiction strategies that span the entire enterprise protecting it and all its customers from current and future threats.
If you’d like to learn more about how Teradata is deploying AI at leading financial services businesses, please get in touch or arrange to meet up Money20/20 conference in Amsterdam this week.