- Client: NDA – a large provider of Roadsite Assistance Services
- Services Provided: Conversational interfaces
Shorten callcenter’s waiting queues with a voicebot agent
Our customer wanted to build a voicebot to shorten waiting queues in their callcenter. At the same time they wanted to give human agents more time to focus on problem-solving, rather than going through a script and collecting generic information. Voicebot and human agent will work hand in hand.

The Challenge
Our customer deals with 600-1000 Roadsite Assistance related calls per day, depending on the time of the year. A large part of the process to deal with these calls is well documented in process flows and pre-defined scripts. Our customer had already built an app to modernize their services and automate the Roadsite Assistance process where possible. Within the app, a customer can report an incident and go through the full process right up to the predicted time of arrival of the towing service if necessary. The voicebot we have built largely follows the same process.
After a thorough analysis of the current call-flow, we drafted a list of key information the bot should collect prior to handing over the call to a human operator:
- Licence plate: this is the unique identifier to make sure the caller is actually a customer of our client.
- Make and model of the car
- Nature of the problem (flat tire, car not starting, accident, …)
- Contact number: the known contact number of the customer in the CRM is not necessarily the number he can be contacted on in the next few hours
- Location (address or POI like parking garage, fuel station, …)
After every step, the bot will ask the customer to confirm the collected information is correct. If this isn’t the case, the customer can correct the bot. If after 3 attemts the customer still indicates the bot didn’t collect the requested information correctly, there’s an automated human handover. At all times during the bot-human conversation, there is the possibility for human handover so that the customer doesn’t end up frustrated without a solution for his problem.

Integrations and human handover
The voicebot we have built has different integrations at different times during the call, to re-use as much as possible the existing technology and information already present at our customer. In the very first step of the call, there is an integration with the CRM. The bot sends the license plate through a webservice to the CRM, and checks whether the license plate is a known and thus insured client. If this is not the case, the call is redirected to a human agent. The make and model of the impacted car is also crosschecked with the information known in the database, and the same goes for the contact number. The bot retrieves the known contact number in the CRM and asks the customer to confirm if this is the contact number he can be reached on for updates with regards to his situation. Lastly for the location, we integrate with a well known Maps API to crosscheck if the location of the customer actually exists and to receive precise GPS coordinates.
The purpose of the bot is to collect this necessary, yet very repetitive for a human operator, information. As soon as this is done, the call is redirected to a human agent. On his screen, the human agent will see what information was successfully collected by the bot and can start the remediation process for that specific customer. Human and bot thus work hand in hand to help customers in need as fast as possible.