In this podcast episode, Lorenz Hänggi, Harald Felgner and Marc Saudan from AXA Switzerland talk about AXA’s digital assistant (Ada). Ada is part of AXA Insurance’s DX strategy. The aim of the strategy is to achieve the best digital experience in the Swiss insurance industry.

On Lorenz Hänggi, Harald Felgner and Marc Saudan

All three interviewees work in AXA’s Interaction and Multiexperience Solutions (IMS) team.

• Lorenz Hänggi is the lead developer behind Ada.
• Harald Felgner is a digital experience designer.
• Marc Saudan is Product Owner of the IMS team.

Why a chatbot?

AXA gained its first chatbot experience in a hackathon. The first chatbot then went live in 2018. This led to the idea of creating a higher-level assistant that could do more than just be integrated into a single process or use case.

Ada’s backstory

The first step was AXA’s glass bot. In this case, the focus was placed on a specific use case. This is a rule-based chatbot that customers can use to report glass damage. This use case was selected because only a few details are required to report glass breakage in order to open a case.

Based on this use case, a chatbot was launched that could be found on the usual customer journey. The chatbot was therefore integrated precisely on the subpage where users normally obtain information about glass damage.

AXA’s digital assistant – Ada

In principle, Ada must be able to answer every question that customers have about AXA insurance. It is available in all four national languages.

The Ada approach is based on a question-and-answer game that leads to the goal in a fairly structured way. Its central element is intent recognition. In contrast to the glass bot, the Ada is intent-based. In other words, it allows customers to freely ask any question they want. This option is not possible with a rule-based bot, where the question and answer options are already predefined.[vc_single_image image=”4519″ img_size=”large”]

Intent-based chatbots or master bots

In order to implement an intent-based chatbot, the most common cases that they have to answer on the phone were worked out with the customer center. These concerns were trained as sentences. However, these were internal to the company and did not necessarily match the external questions received. It was therefore important to the AXA team to get the first chatbot live as early as possible. This enabled them to learn from customers’ questions in order to better define their intents in a short space of time.

Expert Bots

Ada can recognize and categorize the customer’s concerns. However, it is not in a position to answer every question in detail. Their function is to recognize the user’s intention and then forward the specific question to so-called expert bots.

In this process, the issue should be resolved for the user in the same communication. It should not be apparent to the user that Ada’s request is being forwarded to an expert bot. Ada acts as a mouthpiece on the surface, so to speak. Technically, however, the bots are completely separate and can be implemented independently of each other using their own technology.

In summary, it is an interaction with Ada and the expert bots. AXA does not have a bot that can do everything, but one that triages to several experts who then respond to the specific questions.

How is a project like Ada received?

The IMS team is attracting a lot of interest with the chatbot project. In particular, there seems to be a need to implement use cases more quickly. An important principle in AXA’s strategy is that the technology remains open and works with different services. Ada is intended to be a customer interaction platform that can also be controlled by other systems.

A feedback and evaluation system was initially created for interaction with customers. The evaluations showed that if a bot cannot progress on its own, the use of a hand-off function is important. This forwards the customer to a person in the live chat. Customers are often very satisfied with this, as the chatbot continues to help them despite not providing an adequate answer.

It’s best to listen to the podcast episode with Sophie Hundertmark, Lorenz Hänggi, Harald Felgner and Marc Saudan for yourself. Have fun!

And if you don’t want to miss any more podcast episodes, then sign up for the mailing list at the bottom of this page.