Search Engine recommendation tools, customer service… The French Bank is increasing the number of services based on artificial intelligence for its individual and very small business customers.
ARI. This is the little name given to BNP Paribas’ new internal Search Engine ARI actually stands for intelligent research assistant. In concrete terms, it allows advisors to ask questions on more or less complex products to a database of 2,000 answers that grows over time. Approximately 1,000 solicitations are sent to ARI every day. A tool reminiscent of IBM Watson but which the French bank has designed in-house. Our data scientists have developed self-learning algorithms. It’s not a classic decision tree,” says Céline Ansquer, Director of Transformation, Marketing and Digital Retail at BNP Paribas’ French retail bank. The bank in the rue d’Antin is no novice in the field of artificial intelligence. The group, which has identified more than 80 AI projects, has set up a lab dedicated to this technology within its corporate & institutional banking (CIB) branch. More than 180 process automations have been delivered, including an automatic contract analysis system and an intelligent translation engine.
180 process automations have been delivered by the IA Lab of the corporate & institutional banking branch.
In recent months, BNP Paribas has been injecting more and more AI into its retail banking branch, which covers both individual and professional/VSE customers. Today’s customer wants to be more autonomous. Thanks to AI, we can give it more autonomy while offering simple routes, providing quick and personalized answers from a distance,” summarizes Céline Ansquer. Between 60 and 65% of BNP Paribas’ customers (i.e. around 4 million customers) are digital, i.e. they authenticate themselves at least once a month on the website or application. On average, these customers visit their application 15 times a month.
To facilitate this autonomy, BNP Paribas’ innovation teams have developed several chatbots based on automatic natural language processing, a branch of AI that aims to better understand and generate human language. Among them, the chatbot Telmi is able to answer about 480 simple questions such as: how to change my bank card, where to find my bank details… Telmi has been based on the questions most asked by Internet users on the BNP Paribas chat for several years. If he is not able to answer the client, he will automatically redirect the client to an advisor. This bot is accessible on the BNP Paribas application and website but also on Messenger and Twitter. Telmi averages 65,000 customer conversations per month.
A customer service centre doped with AI
BNP Paribas also seeks to be more proactive with its clients. We want them to be able to set their own alerts according to their needs, such as when their salary is paid, when they have received a refund, when a cheque has been cashed, when their balance has gone into the red and so on. But we want to have the right level of notifications so that they are not overwhelmed and that they are useful to them,” says Céline Ansquer. Before launching these new functionalities on BNP Paribas’ My Accounts application, the Group will test the notifications on its Hellobank! online bank at the end of 2019. We are working to enrich the services offered to clients around banking coaching with predictive models: indicating to him that he will receive his salary, his allowance, a reimbursement… This allows us to indicate a forecast balance when due, or to alert him to an unusual payment,” illustrates the manager. Functionalities that would enrich its aggregator, which is currently limited to categorizing expenses. What about the voice? Natural language was the most mature response. We’re starting to work on voice biometrics,” Ansquer replies, without further details.
With the same objective of making clients more autonomous, BNP Paribas has embarked on a major project to be launched in 2020: a Customer Service centre based on artificial intelligence, representing an investment of more than EUR 60 million. The French bank has developed with a technology provider (whose name is not given) a solution based on natural language processing and semantic analysis for incoming flows (phone calls, emails, social networks etc.). Objective: to recognize the customer and the reason for his or her request in order to route him or her to the appropriate advisor as quickly as possible. Our goal is that 100% of our clients’ requests are handled within 24 hours and that 80 to 90% of requests are processed within 24 hours depending on the technicality of the subject,” says Céline Ansquer. A total of 300 people will work for this service centre located in Mérignac. The latter will still be less responsive than the neobanks N26 and Revolut, which have virtual client services accessible 24 hours a day, 7 days a week. The BNP Paribas branch network will also be equipped with these tools in 2020.
Even though the focus is on individuals, BNP Paribas is beginning to address the problems of its professional customers (585,000 professional and VSE customers), who are often neglected by traditional banks and who have made the success of the new BtoB banks Qonto, Shine and Anytime. In particular, she works with OneUp, a young publisher of automated accounting software based on AI. This fintech translates 95% of the bank entries in company accounts. It conducted a pilot with 1,000 professional retail banking clients in the Paris region over a six-month period. When interviewing our professional clients, we observed that they spend more than 10 hours per month on administrative tasks. With OneUp, we allow them to save 30 to 50% of this time, so they can devote it to developing their own business, promises Céline Ansquer. BNP Paribas intends to integrate this tool into its bank for professionals by the first half of 2020.