An intelligent chatbot to assist call centre colleagues


Provide call centre staff a tool to help assist customers faster

Sometimes a transaction on a bank statement has a merchant name that is different from the actual operating name of a business. This can cause customers to panic and assume the transaction is fraudulent. It can equally be confusing to the staff dealing with these customer enquiries if they too aren’t familiar with the merchant name.

My task was to design a tool that colleagues could use to find information about merchants to assist customers faster. To do this, the tool needed to be quick and simple to use whilst providing staff a way to add helpful information about merchants that other colleagues could refer to.


See, listen and learn the thought process of customers and staff to design a smart solution

Understanding a customer’s mindset when finding an unknown transaction

Immersing myself in the call centre environment

The first step was to spend time with customers to listen to how they felt when coming across an unknown transaction on their statement. As expected, most people become alarmed and assume it’s fraud — thus they feel violated and want to resolve the issue as soon as possible.

From the vantage point of sitting in the call centre with colleagues on the frontline, I was able to empathise with the enormous amount of pressure they were under. When it came to unknown transactions, customers were concerned and often impatient, so the tool had to be fast.

Involving staff to help shape the solution

Once I was able to understand and empathise with both types of users, I held small group sessions with call centre staff to do some brainstorming. This gave them the opportunity to put their ideas forward and keep them involved throughout the design process.

Brainstorming and shortlisting a solution

Conversation design is the new wireframe

I proposed several solutions to my scrum team and in the end we chose to focus on balancing business and customer needs with an experimental chatbot tool. The fundamental aim was to help call centre staff to assist customers as quickly as possible. Therefore the tool would need to provide helpful information about merchants quickly. Additionally, it needed to allow staff to additional information about merchants to encourage knowledge sharing; thus reducing (or eliminating) the time spent scouring the Internet for answers.

Wireframing took a back seat for this project as the solution was all about conversation (with a chatbot). It was important then that I worked with the call centre staff to draw out effective conversation flows that followed the logic of typical conversations around unknown merchants.

Alongside conversation flows, my challenge was extended to design and introduce a usable UI and supporting components. I tested and refined these with call centre staff throughout the design process.

Mapping potential conversation flows between user and chatbot
Visualising the conversation flow in a chat interface

Design the chatbot interface and its components

The chatbot interface needed to support reusable components that were quick and easy to use. I came up with multiple versions and had staff help shape the final pieces. It was important that these were well thought out so they could be reused in other chatbots.

Initial Chatbot UI Sketches

Component Ideation

Chatbot Card (API) Ideation


High usage of chatbot but efficiency challenges

Staff were impressed with the speed and quality of information provided by the chatbot. However, challenges began to surface as they were under considerable pressure to take on large call volumes. This left staff little time to contribute to the chatbot’s knowledge base – a fundamental design component to the chatbot’s power.

Overall it was a valuable learning experience for my team and me as it paved the way for future chatbots that will be introduced within the Bank. It too has provided me another tool to add to my UX arsenal: thoughtfully planned conversation design.