We are living in a time when the use of language is no longer restricted to communication between human beings as “autoresponder software”, also known as chatbot
s, has been developed to automate conversations and interact with people through messaging platforms. Today, a chatbot has become a “smart assistant” that every business must have in order to communicate with customers in real time via various social media channels.
But did you know that, just like humans, chatbots come in various degrees of intelligence and abilities depending on how they are programmed and the data inputted? Chatbots can also learn, and the more they learn from their database and from conversations with users, the smarter they become
Chatbot Development at Ngern Tid Lor
Social media has been on the rise, resulting in a 327 percent increase in our customer inquiries
via social media channels in 2019, both general and auto insurance inquiries
. Training more customer service staff to answer customer questions no longer seemed adequate or effective so we started to develop our own chatbot to work as a smart assistant for our staff and a right-hand assistant for our customers.
Ngern Tid Lor’s
chatbot is created and managed by the Customer Service - Non Voice
team, which is responsible for social media customer service, and the UX Designer
team that is in charge of creating positive customer experiences across platforms.
While our CS - Non Voice team compiles questions that have been asked by customers as well as correct answers to be given to customers, the UX Designer team uses the compiled questions and answers to build a dialog flow broken down into conversation topics such as auto insurance, motorcycle loans, branch phone numbers, and monthly payments. This dialog flow not only enables the chatbot to follow conversational paths to find the correct answers that correspond to customer inquiries but also helps customers to purchase the right insurance, or directs them to our branch staff for loan services.
A Focus on Customer Experience
During the process of chatbot development, we place much emphasis on the chatbot's intelligence to ensure that it can respond to every question in addition to designing the conversational style that makes customers feel like they are talking to our staff. This is why we named our chatbot “Jingjai”, because not only can it provide smart customer services but it also reflects our corporate personality traits of approachability and sincerity.
By “intelligence”, we mean that the chatbot must be able to analyze and “extract” what customers really want in their answers. For example, once the chatbot knows that the customer is interested in a car title loan, it will ask which car brand the customer owns. Some customers may answer the brand only; in this case the chatbot
will further ask for the model and year, and so on. But, if the customer tells the chatbot everything in one go, including the make, model, and year, and it still asks about the model and year, the customer will surely get annoyed and will not want to talk any more.
How to Input Data to Make the Chatbot Understand Customers
We need to teach the chatbot about all possible questions that our customers may ask. Other than the questions, we also have to add words that are often misspelled or wordings that are grammatically incorrect or which deviate from standard scripts, so that the chatbot can understand what customers intend to say. This is why the Customer Service - Non Voice team, which is responsible for customer care and answering customer queries, has further developed skills to become a trainer to the chatbot.
In addition, we analyzed conversations between customers and the chatbot
and found that customers often typed long sentences that the chatbot did not understand. We have, thus, designed and implemented the API matching for our chatbot to enable it to interpret the messages and match them with the learned answers with an accuracy of up to 90 percent.
Chatbot Jingjai and Its Impressive Performance
In addition to answering customer questions in real time and forwarding information about customer needs our staff, Jingjai also provided answers about debt payment deferment during the first wave of the COVID-19 epidemic. This allowed our customers to resolve their problems faster because one chatbot can assist 22 customers at any one time. A survey of customer satisfaction also shows that customers were satisfied with the Jingjai chatbot up to 98 percent.
As mentioned above, we attach great importance to the chatbot's intelligence to ensure that it can respond to every question. As a result, our Jingjai chatbot has a complex flow that allows it to deliver great user experience and has been chosen by Facebook as the Facebook Chatbot Best Business Case Study
. This achievement attests to our commitment to developing digital innovation to create a positive experience and deliver the best services to our customers.