Simplifying Bank Processes with Artificial Intelligence

Published Nov-23-19

Breakthrough:
One of Spain's biggest banks develops an AI-based system to simplify direct debit payments.

Company:
La Caixa, Spain

The Story:

Artificial intelligence is a fast-moving technology that is being adopted by numerous industries, making some of their processes better and smarter day by day.

One of the leading adopters of AI is the banking industry, implementing new systems to reduce security risks, add efficiency and transform the customer experience.

In an OpenText survey of financial services professionals, 75% of respondents from banks with over $100 billion in assets said they were planning to deploy AI-enabled solutions.

Experimental System

La Caixa, a leading Spanish retail bank, has been working with AI since 2014. The first commercial application was a smart assistant for foreign trade, and in 2017 it launched the Spanish banking sector's first chatbot. A 2019 update of this system lets the chatbot carry out real transactions.

The bank has also been working on a project that uses artificial intelligence to manage returned direct debit payments. With around 450 million direct debit payments every year, La Caixa is the country's primary direct debit service provider.

The traditional way the bank manages returned direct debits is that software notifies one of the bank's customer advisors when an account has insufficient funds to pay a bill. The advisor looks at the account and decides whether to allow the bill to be paid straight away, postpone payment until there is sufficient balance or reject the bill altogether.

Banking advisors receive information every day at 8 a.m., but the task can be quite laborious and is difficult to automate. This is because numerous variables can determine the action an advisor takes. These include the type of bill, whether it's a regular payment and a customer's payment history.

As the task of managing bills takes up a lot of time, La Caixa decided to invest in an AI project to develop a system whereby the technology can imitate human reason and understand the specific features of each situation. In other words, process each case exactly as a bank employee would.

High Levels of Accuracy

The bank's data scientists developed an AI model that was put through four months of training to 'learn' how to manage a range of different situations. By the end of this training period, the model had achieved an accuracy of 99%, meaning that it nearly always thought and acted like a real bank employee. It was then rolled out to 600 of the bank's almost 6,000 branches. During the first four months of operation, artificial intelligence resolved a high percentage of cases itself.

So successful was the trial project that the bank is considering installing the system in all branches. If it does so, estimates claim that around 82,000 hours of work could be saved. This would allow employees to focus on offering a more personal service to customers.

Next Story »