Artificial Intelligence in banking: you ain't seen nothing yet

For those of us who regularly spend time online, interacting with robots is already part of our everyday lives.

We're accustomed to the Artificial Intelligence (AI) that prompts us to watch another Netflix series, similar to the one we've just finished. We are equally comfortable with the 'chat' window that pops up while we're shopping, even though we know we are interacting with a computer system, not a human being.

These forms of AI are relatively simple, and many rely on technology we have had for 20 years, such as decision trees and algorithms. But AI technology has moved on considerably in recent times. An old-style AI bot may not contain any actual 'intelligence', just a simple ability to mimic language and deal with key words. New AI bots are able to learn about you, thanks to a huge increase in their data-processing capability.

Massive data sets, large computers and very advanced algorithms to churn through the numbers are allowing technology to do amazing things. And this is just the beginning. We are standing on the verge of an exponential increase in AI's use, especially in banks and other financial-services companies.

One recent survey suggests that AI will be the primary way in which banks interact with their customers within three years. Forward-thinking institutions are already launching more sophisticated chatbots. These include Swedbank's Nina, which deals with more than 30,000 queries per month, while RBS is testing Luvo, a bot with a "warm and human-like personality", according to the bank.

These new-generation chatbots can deal with straightforward banking questions, freeing up staff to deal with more complex queries. Our experience at Temenos is that customers don't mind using a bot – and may not even know they are doing so – until something goes wrong.

That is when banks need to have the right systems in place to ensure humans take over as quickly and seamlessly as possible. We know that, once a customer has been disappointed by a chatbot, they won't use that channel again. So if a chatbot is unable to understand a customer's query, or is unable to deal with it, bank staff must be ready to step in.

Chatbots can provide great cost savings, but AI can be embedded deeper still into banks. It is already being used to help prevent online fraud. You may not know it, but bots in the background of your banking apps are analyzing your banking transactions for everything from how well you spell to how fast you swipe, and flagging up anything out of the ordinary as potentially fraudulent.

The same technology can be used to analyze spending and earning patterns to allow the bank to customize products. Many banks already use data to tell customers when they are about to go into overdraft. Smarter banks will use that intelligence to offer them options to deal with it – there may be money in a savings account that could be moved to avoid an expensive overdraft charge, for example, or the bank could pre-approve a loan.

In future, bank bots could run our finances 24/7, constantly monitoring and optimizing the interest rates we pay and the financial services we use. Using AI, banks will be able to learn our preferences and understand our attitudes to risk. It is possible that, before long, they may be able to complete financial transactions for us too – although that is clearly a step that would be fraught with ethical and regulatory concerns.

The advent of new regulations, such as the new European Payments Directive (PSD2), is likely to accelerate the advance of AI. The changes mean that, with customer permission, banks will have to allow outside firms access to account data. Third-party providers, including creators of specialist financial technology, will be able to offer products directly based on the data held by banks. Customers, meanwhile, will be able to access all of their account information, from any bank, from one place.

The winners are likely to be Account Information Service Providers (AISPs), which, with the client's prior permission, will be able to access financial services on their behalf. There will be a vast, real-time online market for financial products to choose from, and AI will play a crucial role in matching the right products to the right customers.

This might involve moving a customer to a credit card with a better rate when one becomes available, or flagging a suitable investment that matches a client's risk profile. AI might even allow service providers to offer instant financial advice based on the type of transactions that the customer is carrying out.

Before this happens, however, regulators, banks and their fintech partners will have to address some inescapable ethical questions that lie at the heart of AI. How much control over our money are we happy to hand over to a set of algorithms – no matter how well they promise to 'learn' our preferences? And what issues will be raised if mysterious black boxes start to deny some people insurance cover or mortgages because the computer says no? How fair and transparent can we expect this brave new world of banking to be?

At present, regulators will not allow algorithms to make serious financial decisions without human agency, and that seems unlikely to change in the near future. However, some people in the industry believe that customers will be willing to part with their data for the sake of convenience and price. So it's possible that one day we will all be happy to allow the bots to bank on our behalf.

Even with strict regulation in place, banking in the era of AI will look very different than it does today. The migration of services from bank branches to mobile platforms, the availability of loans and insurance at the click of a button (or an app), the increasing accuracy of credit risk thanks to improved analytics – all mean that the banks that get it right first will be the ones to gain the most from the march of the robots.

One thing is certain: today's clunky chatbots will be short-lived. In their place will be tomorrow's sleek, financial optimising machines. We will wonder how we ever managed without them.

Todd Winship is Head of Data and Analytics at Temenos.