Customer service chatbot to stop the support tickets
conversation topics (and a human ain't one)
decrease in human time needed
A 'simple' bot to solve an expensive problem
An international publishing brand of a Fortune100 company had a problem. They produced lovely content and email service, but, people kept contacting them with random questions and asking for help using the website.
They're not a product company, so an army of customer service staff answering questions all day was not an option. In fact, for the international sites (not including the USA) they had just one person responding to customer service tickets. That poor fleshy human.
They needed a way to lower the number of tickets hitting the human's desk. However, what they didn't want is to reduce the quality of the service.
A well-trained machine-human was required.
Intercept and serve
The goal was to build a service chatbot to help users before they hit that 'contact us' submit button. It was to be trained from a niche knowledge base, things like "how do I log in?", "I want to unsubscribe" and "why doesn't my app work?". With this in mind, we didn't want the bot to appear on every page of the website. It was designed to intercept and serve the user on the FAQs, contact and complaints sections of the site.
As both a live-chat pop-up and embedded on-page solution, the chatbot reached out to users when they clicked the contact button or dwelled on-page for too long. The goal was to answer their question immediately, before they created a customer service ticket.
It was important the chatbot didn't upset people. After all, not everyone wants to talk to a machine. With this in mind, the chatbot made it very clear how to speak to a human should the user wish to. It contained a triage function where, if they did want to talk to a human, they were asked a couple of quick questions to asses the immediacy of response required.
Real-time service or take a message?
To help lower the service ticket volumes even further, we built a live-chat, human takeover function. The client was able to set 'opening hours' for the bot, times of the day where the bot would facilitate a human-to-human conversation through the chat window. This means the customer service human, when they were at their desk, would be able to help users right away.
Outside of the opening hours, the client could tell the bot to take a message (and gather specific information). It would automatically create a service ticket on Salesforce and prioritise them appropriately, ready for when the human starts work the next day.
Finally, as an excellent customer service team member, the chatbot would follow up with the user a few days later, making sure their problem was solved and checking to see if they needed any more help.
Over the first 14 days of the service chatbot being live, the average daily service tickets fell by almost 48%.
That customer service human got back 48%, almost 4 hours, of their day. She liked the ubisend geeks.
Due to the confidential data we help our clients manage, we often work under hefty NDAs. If you’re interested to know more about this project, get in contact, as we may be able to introduce you to the client for a chat.