(This article was published in Entrepreneur India)
A chat bot is nothing but a mechanism in which humans can interact with a device, which understands the intent of the human and guides them to a desired response. The earliest form of a “chatbot” would have been the IVRS (Interactive Voice Response System) technology that we use on our voice calls. You dial a number, and a recorded voice gives you options, and you select one by punching a number on the phone. And you do this iteratively a few times…and hopefully you get the answer that you are looking for.
A bot is a piece of software that runs in an automated fashion. And a chat bot (also known as conversational bot), gives the human an impression of engaging in a conversation with the machine. The emergence of messengers (Whats App, Facebook Messenger, Slack etc) over the past few years, has made the Internet very conversational. As a matter of fact the user of messengers has surpassed the use of social platforms, globally. Soon, the days of structured communication, like what we were taught in school, that we use in letters and to some extent in email, will be over. And all communication will be chatty and seeking instant gratification (in the form of quick responses). And while using a messenger we can chat with friends and colleagues, using exactly the same kind of interface, we can now chat with “machines”.
What are Chatbots used for
One of the commonest application of a chat bot is in the field of customer service. Chatbots certainly help to reduce costs; remove the limitation of limited number of human agents who can take the call; and also ensure better accuracy in data capture in the conversation.
You want to order a pizza ? You don’t need to call up a help line. Nor do you need to send an email or fill in a form. Just logon to the chat bot of your favorite pizza brand (you should be able to find this on their website or on their Facebook page), and then place the order like you would place on a phone, either by typing in your responses or calling them out (depending on whether the bot is voice enabled or not). The bot will guide you through the entire ordering processing, asking you make choices (type of pizza, size, toppings, add on orders, address, phone number etc) till you have a complete order and have paid for it. What is true for Pizza is also true if you want to order a cab, or book an airline ticket etc, or get information about weather etc.
Some innovative applications of Chatbots
Here are some of the more interesting and innovative applications of chat bots
- Casper : You are having a tough time falling asleep ? And feel like talking to someone. Well you don’t need to wake up your friend, in the middle of the night, to have a chat. You can just use Casper, a chatbot that assists insomniacs. And have a conversation with Casper till you fall asleep
- Edward : Radisson hotels uses a chatbot called Edward to help with online check–in, customer enquiries and complaints. And if your complaint is not fixed to your satisfaction, Edward also knows the right person to whom this then needs to be escalated for human intervention
- MedWhat : Think of MedWhat as the bot equivalent of WebMed. It helps you with medical diagnosis. It will keep asking you questions and basis your answers, ask the next question…till you reach an answer
- U–Report : U-Report is bot set up by UNICEF to reach out to users and poll them on various issues. And using the response, create action plans. This bot does not really converse, but shows up as poll question, which you can quickly respond using your phone. In Liberia, U–Report send out a survey to a very large number of school children about treatment in schools. 86% of the students responded and their response was used to create an action plan focussing on specific areas of improvement in schools in Liberia.
- RoofAI : This bot is used by a real estate company, to understand and assess a potential business lead. The bot asks you a few questions to help understand your requirement better…and then basis your answers, assigns your case to a real human for further engagement
- Tay : This was a bot released by Microsoft to create responses on Twitter. The bot was asked to ‘read’ tweets on Twitter, do some self-learning and basis those tweets, engage with users. However since conversations on Twitter are hugely polarised and very radical, the bot started generating tweets with expletives and radical views. And Tay had to be pulled out
Types of chatbots
Chatbots can be classified into two categories
- Rule based chat bots : The bot is pre-programmed with the entire content. A detailed if-then tree is created, with multiple branches. And basis what in the input from the user, the bot selects the correct branch and responds with that output. And this process happens iteratively till the user gets the information or solution that he is looking for. If the user responds with an input that is not already in one of the branches of the tree, the bot will not be able to respond.
- Machine Learning Bots : These bots, in addition to having, pre-programmed if-then tree, can also create branches to the tree on their own, as they keep learning from the usage. Of course one needs to keep in the mind that even in case of machine-learning systems, if the finally the programmer who is creating the algorithm for the machine to be able to make sense out of a conversation.
While there is an illusion of simplicity on the front end, there are many hurdles that need to be overcome to create a great chatbot experience. Over the years, our minds are used to thinking of communication in the form of web pages with home button, back button and information laid out in neat menu and sub-menu like structure. But bots don’t have back button or home button. And there is only so much information and so many options that a bot can give out in one single interaction. Hence organising the information properly; envisaging all possible user paths to a given piece of information; analytics to keep understanding what is working and what is not; and technology to keep pace with ever changing formats of platforms is needed to create a successful chat bot.
Here are some best practices while designing a user interface for a chat bot :
- Close-ended-answers : Instead of having statements that are open-ended, it is good to put out statements that are close-ended…and provide the user a set of options to choose from at every branch of the tree. This will help intelligently guide the user to the solution
- Personalisation : Keep referring to the user name every once a while. It make the conversation seem more personal. Also use emojis and relevant visuals to keep the conversation interesting, rather than just pure text
- Keep user engaged : If the user wants some information which is taking a few additional seconds to process, provide the user some generic trivia and keep the user engaged…. even while the machine is searching for the requested information in the background