Skip to main content

IA and RPA Saga: Smart Solutions in Banking Customer Services

As part of the work around Artificial Intelligence (AI) and Robotic Process Automation (RPA) in retail banking, Sia Partners offers a saga of three articles offering thematic zooms on banking processes affected by AI and RPA: customer service, regulatory and pre-sales.

In particular, AI and RPA are disrupting offers, paths and banking processes. Optimizing the customer experience is at the heart of the discussions, with major business use cases emerging:

  • The predictive analysis component with, for example, fraud detection (big data technology and analytics),
  • Process optimization such as, for example, linking with biometric elements,
  • Smart assistants with chatbots for instance.

A few figures illustrate the importance of AI: 98% of the companies surveyed believe that process automation is vital to achieving a competitive advantage[1]. However, the implementation of these solutions by these same companies is still in its infancy. Only 15% of them use AI and RPA solutions for their customer service issues[2].

Strategic challenges

First, let's define the concepts of AI and RPA. RPA is a robotic process automation technology[3]. In other words, RPA consists in automating simple and repetitive tasks that require little or no human reflection. For example, it is possible to use RPA technologies to update the tracking data of an Excel table with structured data from another computer program. The leaders in automation solutions are Romanian UiPath, American Automation Anywhere, and English Blue Prism.

The AI allows the robot to adapt to the environment and to propose solutions that are appropriate to the context and the need at a precise moment. The robot is no longer just a simple task executor but also becomes a force for proposals[4]. For example, a self-learning system can be set up to respond to customer requests. Watson, IBM's solution, has established itself as the leader in AI applied to operational issues. The latter is able to recognize words and images, understand language, analyze data and make predictions. AI and RPA are complementary concepts, and represent different degrees of automation.

These technologies enable banks to reduce their operational costs (a robot costs 15% of a French employee4), to reduce errors (the human error rate on a task can reach 10%3), which makes it possible to increase the quality of the service provided, and ultimately, customer satisfaction. RPA also benefits employees by automating low value-added activities and giving them the opportunity to focus on more rewarding tasks.

Multiple applications at all levels of the bank

The AI / RPA solutions implemented by organizations are multiple and at varying degrees of maturity and technological advancement. ANZ Banking Group, a pioneer in the adoption of RPA, has been using automation in several areas since 2015: account reconciliation, monthly closing and customer data maintenance. At the end of March 2016, 500 robots were operational at 8 different sites[5]. LCL chose a more complex solution focused on its credit back office; provided by UiPath, 30 robots were deployed on 17 activities, including simplified and automated processes for examining professional financing requests[6]. In order to reconcile its IT system composed of technologically heterogeneous applications, Natixis opted for Contextor, thus allowing it to reconcile business processes in a more fluid and less time-consuming way for employees[7]. Finally, Société Générale has set up an AI allowing it to detect fraud via a chatbot in test for customers[8].

This first article of our saga now returns to one of the three target themes: customer services. This function manages part of the relationship with clients, its action is strategic for a retail bank.

Use cases related to customer service

Several types of technologies, with different degrees of automation, are used in applications such as authentication, customer dissatisfaction detection and intelligent assistants.Several types of technologies, with different degrees of automation, are used in applications such as authentication, customer dissatisfaction detection and intelligent assistants.

Voice biometrics is an authentication method that has been acclaimed by banks. The underlying algorithms allow you to define a voice identity based on several criteria such as intonation, bit rate or tone. The most emblematic example is that of La Banque Postale, which uses voice recognition to secure online payments in its Talk to Pay application: instead of receiving an SMS, the customer is automatically called and then has to say a previously configured sentence. After this step, he receives a code to validate the payment[9]. In the field of customer services, ANZ Banking Group is testing the use of voice recognition technologies as an authentication method in its call centres[10]. The group uses Nuance, the industry leader. Other players such as United Biometrics also offer solutions to secure access to banks' customer databases.

Another use of AI in customer banking is to detect or anticipate customer dissatisfaction. Many solutions offer semantic analysis software to automatically detect and analyze customer feedback and conversations in real time. Explore, Dictanova or Proxem, evaluate customer satisfaction or dissatisfaction as well as the source associated with it according to the themes addressed, the words used, the tones, etc. These solutions aggregate data from multiple media, allowing monitoring of traditional (e.g. call center) and digital (e.g. social networks) customer expression areas.

Finally, the use of intelligent assistants is certainly the most telling example of the possibilities that AI offers in terms of enhancing customer relations. Chatbots, thanks to the use of RPA, big data or deep learning, allow a simplification, personalization and deepening of customer management: savings / insurance diagnosis, preparation of appointments, writing of interview reports, processing of complaint emails (and proposal of answers) or even chat (conducted in natural language) to satisfy customer requests, their role is variable in size. From a simple personal assistant (the counsellor, assisted by a chatbot, becomes an "augmented" counsellor) to a real proactive collaborator, setting up predictive models to anticipate the client's important life moments (e.g. unemployment, birth, death, etc.) and act accordingly, chatbots become essential.

The example of Djingo, the virtual advisor of Orange Bank

When Orange Bank was launched last November, the bank created a real breakthrough in customer relationship management by entrusting the so-called first-level requests (e.g. contact entry, card requests) to a virtual advisor called Djingo.

Developed on the basis of the Watson artificial intelligence solution (designed by IBM), Djingo responds to customer requests (more than 500 questions can be answered in natural language, imitating human tonality) and performs low value-added operations, such as blocking / unblocking a credit card. The assistant conducts an average of nearly 24,000 conversations per week (for a total of more than 100,000 conversations per month) with an understanding rate of 85%. On the other hand, Djingo does not intervene on subjects identified by the bank as complex and/or primarily related to human relations (such as the contestation of a sample for example). In these particular cases, the virtual assistant quickly passes the baton to one of the 300 human call centre agents, who then access the conversation history.

Based on a model of continuous learning, AI progresses gradually, relieving humans of ungrateful and time-consuming tasks. As proof, Orange Bank has begun to test an email analyzer (already used at Crédit Mutuel): in 75% of cases the request and its degree of urgency are successfully identified. Finally, Djingo is expected to become proactive this year: he will soon contact customers on his own initiative, to propose to them himself (and in complete autonomy) actions that meet their needs.

An emerging trend that banks cannot avoid

Whether as a means of simplifying, reducing stress or improving customer relations / satisfaction, the fields of application are not limited to front-office activities but are on the contrary multiple: regulatory and pre-sales compliance in particular.

The next article in the saga will deal with regulatory issues, with a focus on the processes impacted by RPA and AI.



Notes & References:

[1] CIO Insight, 2014

[2] Harvard Business Review, 2017

[3] Journal Du Net, 2016

[4] Blog CIO Advisory Sia Partners, 2017

[5] Hello Finance, 2017

[6] Le Mag IT, 2018

[7] Les Echos, 2017

[8] Capital, 2018

[9] La Banque Postale, 2017

[10] Banking Frontiers, 2017

If you want to know more about our AI solutions, check out our Heka website: