The Client

ING is a Dutch multinational banking and financial services corporation headquartered in Amsterdam. Its primary businesses are retail banking, direct banking, commercial banking, investment banking, wholesale banking, private banking, asset management, and insurance services.

The Search & Intelligent Assistance team is responsible for the search functionality of the Home and Business applications for the end customers in the Netherlands and Belgium. It provides all the frontend views and backend infrastructure to help customers in finding the relevant information, tools, and products in the Home and Business banking applications. Additionally, it provides conversational assistants to guide users to relevant pages, resources, and products.

The Project

The purpose of Search is to guide users to the most relevant available resources; therefore it has moved on from simple documents search towards a complex set of tools able to search and display transactions and other personalized data, and to detect customer intent, displaying wizards to guide them through processes (for example, Credit Card activation). The last addition to the available set of tools is a full-fledged AI Chatbot helping users in an interactive way in navigating through the available resources and product options.

In this context, the project revolves around the need to improve the current Chatbot solution, extending its capabilities and integrating it better with the existing flows and interfaces. For the backend and development part, adding new chat flows to the existing APIs requires a developer intervention, and maintaining their configuration in the external AI tool used to drive the conversations (Google’s Dialogflow) is extremely complex and error-prone. On the frontend, various experiments need to be performed to decide the best way of integrating the new functionality in the existing flows or of adding it to flows outside the Search page.

Technologies

  • Polymer 2, Web Components, Vue.js
  • Nodejs, Typescript, Java
  • Postgresql
  • Google Dialogflow

Using the existing methods to implement and deploy new chatbot flows was impractical, requiring a large amount of backend work and of manual configuration of the chatbot engine. Managing complex conversations proved to be almost impossible due to the complexity and unfriendliness of the setup. On the frontend part, it became obvious that there was a need to integrate the conversational agent seamlessly into the page, allowing it to interact with the page itself and to respond to the actions performed by the user; this prompted a move from the traditional chat window to an agent directly appearing inside the current page flow. This was obtained through the unification of some of the existing interfaces to a conversational style and a significant functional and visual improvement of the existing Intelligent Assistant interface.

Additionally, Sytac proposed and implemented a new software tool to allow Customer Journey Experts to visually develop, test, and deploy complex conversational flows to guide customers through a large number of products and options, by using a simple YML configuration schema and a graphical representation of the flow. On the frontend, we worked to extend the capabilities of the chat agent allowing it to interact with other components on the page, to respond to user actions performed on them, and to display external content (starting with a search box and a feedback tool, but able to also display, for example, search results). Each frontend change was extensively A/B tested to check its usefulness for the customers.

Results

For the business, we provided the ability to design, manage, test and deploy complex conversation flows that could be integrated anywhere in the banking application, bringing the benefits of Search & Assistance to other teams, all with less involvement from developers. Conversions were increased in several scenarios after the introduction of the agent.

For customers, we helped in providing a unified, easily recognizable interface for all the Intelligent Assistance communications inside and outside the Search page. The assistant helps navigating pages with multiple options, also allowing to explore different products and options.

  • +18% increase in loan sales
  • 13k customers/ week redirected to the correct location
  • 250 user feedbacks collected per week