[Testimonials] The deployment of SAM, Banque Populaire's chatbot

Department Head and Project Manager at Banque Populaire Grand Ouest Marie-Annick Mevel and Chatbot Manager Julien Bichon at Informatique Banque Populaire were in charge of the deployment of the virtual assistant SAM to be used by the firm’s consultants. This virtual agent has made it possible to provide users with first-level support, relieving specialized experts of the necessity of addressing frequently asked questions. Marie-Annick Mevel and Julien Bichon agreed to take part in a Q&A session to tell us all about the implementation and operation of SAM!



How did you decide to implement a chatbot at Banque Populaire?

M.-A. Mevel: The professional and regulatory nature of the banking and insurance industry is constantly changing. For this reason, consultants are increasingly calling on specialized professional expertise. Now, the recurrence of certain questions, the waiting times for obtaining responses and the informal requests between colleagues prompted the IT Department to reconfigure internal assistance.
The challenge was to reduce the workload of the specialized experts in order for them to focus on complex client cases and concentrate on problems with more value added. In response to that challenge, we opted for a chatbot.
J. Bichon: The technology was already in place, as Living Actor was already working with the BPCE Group. Our virtual assistant, christened SAM, enabled us to lighten the load of the assistance network by adding a layer of first-level support. This chatbot handles consultants’ frequently asked questions regarding regulations, professions, job positions, procedures, etc.

Read also: Are Chatbots the Future of Banking?

M.-A. Mevel: Basically, SAM functions like a security vestibule. Consultants first present their inquiries to our chatbot. After that, there are two possible outcomes:

  • they’ll either receive a relevant response, saving our experts time,
  • or if the inquiry is unresolved, SAM will offer to put them in touch with the person who will most likely be able to assist them.

We no longer provide the specialized experts’ direct contact information: it’s mandatory to go through the virtual assistant first.
SAM’s other advantages include availability and discretion. On the one hand, the bot can respond to inquiries when the experts aren’t around, for example, on a Saturday morning. On the other hand, during a consultation with a client, the consultant’s request is much more discreet than, say, a telephone call would be.

Who is SAM?

J. Bichon: SAM is our virtual assistant. And by the way, SAM is female! We held workshops with two banks in order to determine her appearance, her voice and her tone… in short, her editorial charter. Then, we defined the rules to ensure that SAM would respond relevantly and effectively to the questions asked by the consultants.

Read also: 3 Criteria for a Successful Chatbot Conversation.

How does she work?

J. Bichon: SAM operates on a shared knowledge base, a sort of common trunk for all of the Banque Populaire branches. Each establishment also avails of a smaller, proprietary knowledge base, consisting of personalized information that is unique to that particular location.
In this way, SAM is equipped to respond to general inquiries, as well as those related to regional specifics. As a result, the knowledge base is continuously enriched, as the writers (who prepare the questions and responses that are built in to the chatbot) can capitalize on the information drawn from the local establishments.
M.-A. Mevel: We’ve observed that 80 percent of the inquiries made by our consultants are in reference to 20 percent of the total volume of knowledge. Thanks to SAM, if one consultant wants to learn about the regulatory framework for a single-family home construction contract and another one simply wants to find out how to switch a keyboard to AZERTY layout, both of them will find their answers in the same database, regardless of which establishment they work at.
Interaction with the virtual assistant is instantaneous and pertinent. As a result, SAM enables the consultants to feel more at ease when they need to respond to clients or prospects.

How was SAM received by the personnel?

M.-A. Mevel: Specialized experts and users, alike, all welcomed SAM with open arms. The first group, which was responsible for contributing to the chatbot’s knowledge base, had access to a collaborative space, which built on their previous training and provided them with information on best practices. They drafted the content with the support of advice regarding writing, formatting, empathy, keywords, etc. Afterwards, they used a verification process to validate the content against the editorial charter.
The second group was involved early on in the chatbot implementation process. Tests were conducted using small talk to acclimate them to communicating with SAM. We had to encourage them to assimilate the tool and go through the virtual assistant naturally before consulting a specialized expert. That was the entire challenge of our project! The consultants quickly became very receptive and interested.

What role did innovation managers play in this project?

J. Bichon: Our need was born out of a process that had been implemented in a regional establishment. Following an audit, the on-site team had observed something: the consultants were having difficulty knowing which expert to call on regarding their problem. In response, the IT Department had integrated “click-to-call” and “click-to-mail” buttons in the intranet, directly linked to the corresponding experts.
It was a good idea, but it had an unintended consequence: the experts soon found themselves overloaded with requests! That’s when we were contacted at the Innovation Department. We needed to find a solution to lighten the load of the specialized experts.
After comparing the different solutions that were available on the market, we quickly turned our attention to Living Actor technology. It was already being used by the Caisse d’Épargne Group. The initial experimentation with Living Actor was carried out on a small scale: one establishment in late 2016, followed by another in early 2017. Innovation very quickly led to industrialization, and we deployed SAM at almost all of our banks.

Read also: CloE, the Caisse d’Épargne Group’s Virtual Banking Assistant

What is your vision of the chatbot?

M.-A. Mevel: I appreciate the possibilities offered by the bot, while at the same time, I’m aware of its limitations. A virtual assistant that can take the place of a consultant or specialized expert doesn’t exist now and isn’t on the immediate horizon, either. A chatbot’s true intelligence lies in its ability to analyze the questions it’s asked and provide the most relevant responses.
Imagine that you’re a client and you ask the chatbot at your bank if it knows Siri (Apple’s voice-user interface). It will respond with some pleasantry that’s certainly amusing, but also preprogrammed. However, if you ask it how to open a life insurance contract, it will reply that this is outside its field of expertise and that it’s still in the process of learning.
What you need to understand from this example is that the bot doesn’t work without the knowledge previously supplied to it by humans. True artificial intelligence has nothing to do with what’s shown in certain media, where some imagine that machines are already taking over!
How it does work is through supervised learning, in which the expert loads the question and the corresponding answer in a knowledge base. The chatbot’s conversation relies on a succession of scripts. It can’t invent a response outside of this preprogrammed database.

Read also: The Chatbot Manager: Your Virtual Assistant’s Indispensable Coach!

What has been the ROI for this chatbot project?

J. Bichon: After the deployment of SAM in the pilot banks, we observed a nearly 20 percent decrease in inquiries made to the specialized experts, both by telephone and by email. And the time the experts saved was put to good use.
We have also progressively adapted our virtual assistant to users’ needs, thanks to the valuable data collected by the Living Actor platform. This data allows us to evaluate the quality of the interactions and the satisfaction of the consultants. We can also see if an inquiry hasn’t received a response and then ask the experts to draft one!

What are your future projects?

J. Bichon: Our next project is the increasingly extensive adaptation of the information system to users’ needs and the numerous sources of knowledge. The goal is to provide consultants with a unique entry point to support and research through the chatbot.

Read also: 7 Questions to Ask Yourself before Launching a Chatbot

As the innovation managers, the chatbot is a veritable goldmine.For Marie-Annick Mevel, this solution, combining technology and adaptability and implying a significant return on investment, allows to respond as fully as possible to the challenges of the digital transformation of our companies.