Xomnia and the development team in bol.com have collaborated to build an end-to-end web service that can autonomously handle customer queries on bol.com’s website. Using this service, the e-commerce platform will provide a faster and more intuitive experience for its customers, while reducing the costs of their customer service operations.
Case: Need to automate customer service
Bol.com is the leading online retailer in the Netherlands, delivering value to its customers by developing web and mobile applications that embrace technological and business innovations.
With tens of thousands of selling partners operating on its e-commerce platform, however, bol.com has been faced with a laborious and time consuming task: To manually provide customers with the right contact channels to answer their inquiries. During a call or chat, the customer service advisor needs to help customers and use internal tools and services to redirect them to the appropriate 3rd party and inform them about their contact options.
To make this process more time efficient and cost-effective, bol.com partnered with Xomnia to automate responding to customer inquiries as much as possible. One of our data engineers collaborated with bol.com’s development team to provide engineering knowledge and extra capacity to implement the solution.
Solution: A customer service chatbot
The team developed a web service application to automate a customer’s interactions for a wide range of customer service-related questions. In a chatbot-like manner, and based on a variety of user inputs, a customer is provided the best contact channels to answer their question. This way, customers will know whether bol.com or any of its selling partners can best be addressed to answer their question without having to interact with bol.com’s customer service experts from the start.
The application was written in Kotlin and the service was based on the Java Spring web application framework. It runs on the Google Cloud Platform on a Kubernetes cluster that simplifies scaling considerations. At the core of the application, there is a custom implementation of a Finite State Machine (FSM) that enables control of the process of the different stages of a chat with a customer. This determines which would be the most suitable next action depending on the information provided by the user, and the data retrieved from other bol.com services. The information needed by the FSM is persisted temporarily in a Postgres database, whereas data needed for analytical purposes are stored in Google Big Query.
Automating this process – so far a manual, time-consuming task which lacked sufficient granularity – helped bol.com provide a faster and more intuitive experience for its customers while cutting costs for their customer service operations.
Additionally, this project follows the strategic direction of the online retailer to increase the role of AI and conversational solutions, intent recognition, speech and conversational analytics, Neural Network-based chatbots, and others within the customer service department. All these data-based developments will help bol.com usher in a new era of customer service and the customer journey.