The Chatbot Tsunami has brought us a flood of chatbots to help automate various business functions, including customer service chatbots that automate customer Q&A, marketing research chatbots that automate customer interviews, and HR chatbots that automate job interviews. Although a chatbot is often made to serve one purpose, users expect the chatbot to perform multiple related functions while serving the main purpose.
As indicated by this Forbes article, 76% of customers contact businesses to make inquiries and get support via text messaging. Now with the required social distancing during the COVID-19 pandemic, more and more consumers choose to interact with brands online.
Since one of the most popular chatbot uses is to automate customer service, a chatbot's abilities to answer user questions is directly related to brand image and customer satisfaction. Previously, I talked about how to create a Q&A chatbot in a few minutes to answer users' free-text questions and handle complex, multi-turn Q&As, all without coding. Because no chatbot is perfect, I also mentioned how to teach a chatbot handle unknown user questions and further improve customer satisfaction.
In my Clojure/north 2020 talk on "diffing-based software architecture patterns", I mentioned that Juji is using Editscript, a library I developed, to diff Clojure data structures. During the Q&A session of the talk, someone brought up another Clojure diff library, called deep-diff2, which I was unaware of. Then on Youtube, a comment asking the difference between Editscript and deep-diff2 appeared again. This prompted me to do an investigation on Clojure data diff libraries. Given how the Clojure community places such an emphasis on data oriented programming, a comparison of data diff alternatives appears to be of interest.
A while back, I wrote a blog on "Chatbot Tsunami", mentioning that our world is now inundated with chatbots, including both "good ones" that can help us fulfill our requests and "bad ones" that can hardly understand users or achieve anything.
One of the most intuitive uses of chatbots is perhaps to automate customer service, where a chatbot automatically answers user questions, 24x7. In fact, in almost every chatbot application, a chatbot's ability to answer user questions, especially free-text questions, is always desirable.
Students have difficulty receiving career advice due to the low advisor-to-student ratio. I know that's true for me. Sometimes I would forgot to make an appointment with a counselor, dropped-in, and waited an hour to see one. Other times I avoided the office entirely because I knew appointments were full. How could such issues be solved? Maybe it's possible to create a chatbot that could act as a career counselor.
During this difficult time of COVID-19 pandemic, healthcare organizations and medical institutions are inundated with requests and demands from the general public. As an AI chatbot company based in Silicon Valley, Juji would like to lend them a helping hand. We are currently offering our chatbot services free-of-charge to these institutions.
We all know that user interface (UI) development is an iterative process. It is important that we can iterate quickly based on user feedbacks. At Juji, we have been constantly searching for solutions that enable faster iterations for our Juji Studio product. Around the end of last year, we did a major revamp of Juji Studio UI. By all accounts, this change made a huge difference in term of usability of Juji Studio. More importantly, we can now iterate much faster than previously possible. What's more, we did the wholesale changes in less than one month! Here is how we did it.
When people first use Juji, they are often amazed by how easy it is to create an intelligent chatbot with the platform. This reaction of pleasant surprise is particularly pronounced for people in the know, i.e. technical people who have actually done relevant work before. I am talking about the CTOs, the NLP researchers, and the employees of big technology firms.
