Let’s talk about chatbots
This story was originally published in Turn The Page #62, November 2017
Without a doubt you have come in contact with some form of chatbot in recent history. A chatbot is a computer interface that is built around a conversation. They are available 24/7 and are a great way to quickly reply to simple questions, even if there is no human ‘colleague’ available. Chatbots sometimes serve a single purpose, while others are fully fledged digital assistants. Do you wonder why chatbots started popping up all of a sudden? Why should you even care?
Previously, chatbots scanned for keywords within an input. Those chatbots would then pull a reply from a database with the most matching keywords. However, the past few years have seen a lot of development in Natural Language Processing (NLP), a term that encompasses all technologies that allow a computer to understand human (natural) language. Recent developments in NLP are mostly driven by Machine Learning. Machine Learning is based around the idea that we give a computer a lot of data to process, which it then uses to learn skills, like recognising images and natural language, by itself. Natural Language Processing has passed a threshold that allows great chatbots to exist. Big companies like Google, IBM as well as Facebook have developed their own platforms to enable companies to deploy a chatbot relatively easily. The benefits of using a chatbot are numerous, so the explosive growth of chatbots in the past year is understandable.
One major use case for chatbots is improving the customer’s experience with a company. On top of the benefit of being available 24/7 at reduced cost, a well implemented chatbot allows a company to reach a massive amount of customers on an individual level. They offer great potential for personalised communication, like helping customers find recipes (Allerhande), telling them what to wear (H&M) or simply solving an issue they are having with your service (AirBNB, Spotify, et cetera). The goal of these chatbots is to quickly provide customers with answers. These chatbots have a specific purpose, often built by a single company. However, there are general purpose chatbots as well. Google Assistant and Apple’s Siri are well known examples. Google even allows apps to extend Assistant’s capabilities by giving programmers the ability to add features to their app, arguably making it the most versatile chatbot to date.
It is very likely you have come across a customer support line that makes you go through a dreadful interactive menu before letting you talk to a support agent. Users might get frustrated in a similar fashion if your chatbot throws them into a never ending loop of horror. While making use of a chatbot can certainly reduce the size of a customer service department, it will not be able to replace it entirely (for now). If a user feels their question has not been properly answered, he should be able to talk to a human operator instead. Another downside is the obvious acceptance issue, as with any innovative idea. While you might certainly please a young target audience with a chatbot, older generations will not be so easy to convince of the benefits.
I think chatbots will replace many mobile apps in the foreseeable future. Chatbots make it easy and fast to reach customer service in messaging apps you are already using. Even if older people would rather talk to customer service, they would probably prefer it over using an app or website. People understand text well, because it is less abstract than websites and apps. Furthermore, most customer service departments try to reduce the average handling time, as it decreases cost significantly. Chatbots allow customers to talk to a company more often and longer. I think it is a great idea to encourage this, it helps to build a relationship and trust, all while keeping additional costs to a minimum. I am personally really excited about the potential of chatbots, so I have looked into the basics of building one myself.
If you are as inspired by the applications of chatbots and want to build your own, here are some pointers to get you started. First of all, there are a lot of chat clients that allow bots, Facebook Messenger and Telegram are probably the most-used platforms because they are so easy to work with. Secondly you will have to decide whether you are going to hire developers to build your chatbot or if you are going to use a framework, like Google’s API.AI, Chatfuel or Botsify. Those will go a long way, providing an easy to use web interface, requiring little to no coding skills.
Before you start building, think about the structure of your chatbot. Develop a conversation flow that makes sense.
Chatbot Top Tips
- Welcome the user
Tell the user what your bot is called and tell the user what the scope is. What can your bot do? You may also set quick-replies, to guide the user. - Anticipate questions
Think of the questions your user will ask. How will you reply to them? What kind of response does your user expect? Think about rewarding them for using your chatbot. - Fallbacks
If you can not help the user with their question, suggest to ask another question. Also ask them if they would like to talk to a real person if they are not satisfied with the reply. - Log conversations
Make sure your chatbot learns from its mistakes. Register if the user was happy with their response. If the conversation was stopped mid-question, register why. If a question could not be answered properly, register what they asked. Does this question get asked more often?
Getting started is easy, a basic bot can be set up in just a few hours and could be of great benefit to your customers. Start somewhere, work on the core functionality of your chatbot and expand from there, improving your chatbot’s usefulness over time.
