At ERGO, we have been experimenting with voice technology for years. For example, we had the first insurance policy that be bought purely by voice - and now, we have several awards in our showcase. But one application - which at first glance does not look very prizeworthy - is just beginning to glow: We are combining one of the oldest communication media in the world with voice technology.
Be it via live chat, social media, e-mail or personally - a modern service must be able to serve customers on many channels. Mail and chat in particular are becoming increasingly popular with younger target groups. And yet for many customers, a medium that is almost 160 years old is still the first choice: The ERGO hotlines receive 16,000 calls a day. People love to make phone calls - probably because it is more personal and direct.
I also like to pick up the phone - whether in the office or privately with my telecommunications provider. Although routing via chat bot has now reached a reasonable level, the telephone still promises to bring me quickly to my solution. The problem begins when the automatic hotline voice kindly explains that it will take longer today.
This is exactly where we at ERGO dock our smart interactive voice response technology: If customers are willing to talk to an artificial customer advisor, we offer increasingly fluent, case-closing dialogs with dynamically generated speech instead of the usual multi-frequency procedures with "then say contract" or "press 2".
The bot responds to the customer's request, who says, for example, "I am reporting a damage to the car", with: "Is this your car or another vehicle?" This creates dialogs with ten or more interactions. We also connect data interfaces via microservices so that the machine can pull all existing information on the telephone incident from the back-end systems in real time - and then incorporate it into the dialog.
But do people even want to talk to a voice computer? Yes! 60 percent of those surveyed would pick a voice assistant over a human being on the phone. This is the result of a recent survey conducted as part of our ERGO Risk Report.
Another example: We at ERGO have a bot in live operation that helps customers in the event of hail damage. This doesn't happen often, but when it does, many policyholders are affected at once -- and our hotline runs hot. Long waiting times are thus programmed, unfortunately - unless I agree to a virtual service consultant.
The dialogue with the machine starts directly without waiting times. And: It is case-closing! In the end, the damage report is being recorded and forwarded to the entrance management. There, the next process robot is waiting to be deployed. But I will come back to this - to Robotic Process Automation - separately in a later blog entry.
Another application has beend deployed in our health insurance company DKV. For the time being, people with private health insurance pay their medical bills themselves. Then they submit them to the insurance company for reimbursement. Many customers call to inquire about the current processing status. This is an ideal environment for the Phonebot, as such calls are repetitive: Just give your name or customer number and the virtual voice tells you how far the processing is.
We develop and train the language applications on a platform that is directly connected to established language technologies such as TTS and SST from Google. This is an advantage, because this way our dialogues are interpreted completely anonymously via an automated speech recognition and a speech understanding AI.
With each application we learn more and more - and thus can make dialogs more and more "fluent". However, the design process, carried out by our "Voice & Conversation" department, is complex. The software alone is trained on almost 900 terms with which a customer could enter into a dialogue on his contract. Coding 71, for example, is "contract extension", coding "869" is "bank draft".
We create the blueprints in the "Parloa" dialog management system from our strategic partner Future of Voice - or rather, this is where the concrete business logic for handling customer inquiries is being implemented. The end result is a robust, digital assistant. Its knowledge only needs to be trained once in order to enable a consistent user experience across all communication channels. The customer can rely on the same level of quality and brand personality everywhere.
If you want "personal" contact, a Phonebot is not the answer. But if you want to solve your problem without waiting, we have a fast and reliable solution for you. Important to us: The Phonebot is only an additional service option. A jump to the next (human) customer service employee is possible at any time, in case of problems with the speech intelligibility.
Besides reducing waiting times, our speech assistance solutions have another task: They are designed to give (human) service employees the freedom they need for customer conversations with a higher consulting effort. But the truth is that many customers are still unfamiliar despite functioning BOT dialogs. We humans are not yet too used to talking on the phone with an AI.
We didn't have telephony in mind when we launched "Voice" in 2016. In the meantime, we have reached a completely new dimension with the Phonebots, even if we are still at the beginning. Which makes us especially happy and reassured. We are increasingly being asked to make our construction plans for the dialog applications available to third parties as "SaaS" ("Software as a Service"). The interest in the combination of phone and bot is huge!
The products are also in demand within the ERGO Group globally. We are currently developing solutions for the markets in India and Spain. We are also pleased to receive another award: our "multi-channel Conversational AI platform for speech assistance solutions" won second place in the "Project" category of the Digital Leadership Award. Congratulations to the team!
Text: Mark Klein