In order to increase innovative capacity, health insurer VGZ has deployed an innovation team between IT and the business since 2012. Stefan Visser (data catalyst and innovation architect) and Niels Scheurleer (project manager) talk about the working method of this team and about the most recent innovation project: voicebot Sam. Sam now handles over a thousand calls a week. The next step: linking the voicebot to voice routing.

 

The innovation team of VGZ explores application possibilities of new technology. Together with the business, the team sets up experiments based on concrete and relevant business issues. Stefan Visser: “Initially we did this with external parties, but from 2016 the team was expanded with a project manager, business analysts and developers, so that we can also develop ourselves. Last year, the team was further expanded with data scientists.”
Niels Scheurleer: “The team now consists of almost 20 people. Sometimes we use external capacity – for example if we need developers with specific skills. We also work together with specialists such as Telecats.”

Which technology is hot?

Scheurleer explains why an innovation team can work well in a large organization. “The business often has an idea of ​​which technology applications are 'hot'. But translating this into concrete applications is difficult. In addition, the issues of the day in IT and the business ensure that there is little or no time left to think about innovation. Often there are still all kinds of pending matters that take precedence.”

Visser: “We also see it as our role to cut through technology hypes. Furthermore, the IT function is focused on continuity – ensuring that everything continues to work. They were therefore seen by the organization as reactive. A proactive role has been added to this with the innovation team. This also prevents the business from opting for solutions, where IT is expected to 'implement it'. That is not conducive to a stable IT architecture.”

Not immediate financial pressure

What is unique is that VGZ's innovation team has a large budget at its disposal. “As a result, innovations do not have a direct impact on the business budget,” says Visser. “That makes it easier to get the business on board. We can also quickly deliver a prototype. In this way, the business can try out new applications without investments. Of course, we are expected to deliver added value.” Only when an innovation is fully transferred does it enter the budget of the business unit.

Innovations in customer contact are often aimed at cost savings. Demonstrating added value is also sometimes difficult. In the case of non-conclusive business cases, the vision on the technology, for example, can then be the deciding factor in continuing, according to the two gentlemen. In addition, you have to battle legacy, caused by technology choices from the past. For example, good access to data is a precondition for many innovations in customer contact. Visser explains that VGZ's innovation team can be successful because the IT landscape is generally in good order: “There is an extensive layer of middleware including a good API platform with which we can access information from back-office systems. In addition, we develop exclusively cloud-native solutions, where you initially do not have to take much account of the rules that apply to the regular IT landscape.”

Step 1: the chatbot

On to the chatbot. The choice of a chatbot, five years ago now, is partly the result of two other considerations of VGZ. The health insurer wanted to increase accessibility outside office hours and stop using the laborious e-mail channel. VGZ first sought cooperation with IPSoft – known for Amelia – as an innovation partner. In the end, VGZ decided to stop this, because development, scalability and management required too large investments in resources.
Visser: “Our developers then built a completely new chatbot platform themselves. After half a year, we put the chatbot into production and then handed it over to a new team for management and further development.”

Some conversations cannot be automated

As said, sometimes the business case turns out differently than expected. VGZ's chatbot led the way NOT ALLOWED to a reduction in the telephony channel, says Visser. Nevertheless, the management wanted to continue. Visser about the motive: “We have four million customers and on average every customer calls once a year. About 60 percent of these contacts are relatively simple in nature. We would prefer to automate the questions that add little or no value and make them part of a 24×7 service, provided that customer satisfaction is maintained. We also deal with large amounts of complicated questions – think of people who have been to several hospitals and ask for a second opinion. In these kinds of contacts, empathy is necessary and automation is not a solution.”

“In the dialogue with the chatbot you can take doubtful cases into account: when it becomes difficult for the chatbot, the conversation must be passed on to an employee. You should also consider using a chatbot as a learning path,” says Scheurleer. “When customers ask on the phone about things that can also be found on the website, there may very well be another question behind that seemingly simple question. You have to deal with this carefully.”

Step 2: from chatbot to voicebot

VGZ's chatbot is now available a success. In 2020, 582.000 customer conversations were completed, almost three times as many as in 2019. And the chatbot gets an NPS of +46 from customers. The success of the chatbot got the innovation team thinking: why not use the bot within the telephony channel? The idea for the voicebot was born.

Telecats was already on board at VGZ as part of a business initiative to conduct an experiment on speech routing – which involves intent recognition based on the conversion of 'speech to text'. That pilot coincided with the reorganization of the customer service teams, namely in self-organizing circles focused on specific topics. “To make our voicebot work, we needed transcriptions of spoken text and a speech generator that converts text into speech. How do you answer a telephone customer question with artificial speech? For example, from the innovation team we are with Telecats engaged in conversation.”

Personalized service via voicebot Sam

The voicebot called Sam now handles about a thousand calls a week. Sam speaks based on technology of Read speaker, a partner of Telecats. In addition, Sam, just like an employee, is able to offer personalized services on a number of topics. Scheurleer explains: “We know who the customer is, because we ask for the customer number and we perform authentication by sending a code to the customer's phone. When Sam, our digital assistant, takes the call, he already knows what it's about and who he's talking to.”

The subject is currently still determined via the IVR, which is something VGZ would rather get rid of. Visser: “We want to direct the customer to the right solution as quickly as possible. Of open question speech recognition intelligent routing becomes possible. This means that, depending on the customer and customer demand, you can determine whether you should have a conversation handled by the voicebot or better by an employee.”

IT legacy

In doing so, the innovation team encounters a form of legacy, Scheurleer explains. “The choice of the telephony platform was based on the idea that telephony would no longer play a role ten years later. Now we know that telephony and live chat are and will remain important for our customers. Because VGZ's telephony environment is not yet in the cloud, the link with voice routing is complex.” Or in other words: “It is difficult to bring cloud to on premise. Conversely, it provides a nice playing field for innovation.” The possibility of voice routing thus confronts the organization with legacy.

There is also another reason why a migration to the cloud is not an easy option. VGZ works with medical data. In addition, the organization also carries out health insurance for the SZVK, the compulsory health insurance for military personnel. “So our systems must be well boarded up. That sometimes makes technology choices a bit more conservative. And that hinders innovation in customer contact, for example,” says Scheurleer. That is why VGZ has now opted in the first instance to virtualize the telephony platform, which should make it easier to create links with cloud applications.

Open question speech recognition

Visser explains that the voicebot is active in two different areas of application: information provision and changes. “In the second category, customers can now ask questions and make concrete arrangements regarding the health insurance card, payment arrangements and payment of the deductible in instalments. With VGZ's API platform, Sam can check back office systems for a question about an invoice whether there is a payment arrears and what the cause is; but he can also immediately propose a payment arrangement and confirm this to the customer by e-mail. Open-ended speech recognition should significantly improve intent determination (“tell us what you are calling us about”) and contact routing,” says Visser.

VGZ takes into account that customers have to get used to this way of interacting and that dealing with a voicebot is not easy or self-evident for everyone. During these interactions, VGZ asks customers for the Customer Effort Score (CES – the lower, the less effort) and Sam's CES now averages a 2 on a scale of 1 to 5, better than the regular chatbot that is around the 2,5 scores.

About VGZ Coöperatie VGZ implements the health insurance for different brands. The main brand is VGZ. In addition, there are nine other brands and labels: Bewuzt, IZA, IZZ, UMC, UnitedConsumers, Univé, Zekur, SZVK and Zorgzaam. In total, the cooperative insures approximately 4 million customers. VGZ has more than 1.800 employees. A turnover of almost 12 billion euros is realized (2019).

(Ziptone/Erik Bouwer)

Want to know more about chatbots? View Ziptone's Chatbot Dossier

Case, Featured, Technology concentric, Concentrix Catalyst, Telecats