Three types of chatbot

IBM Watson, Mitch Mason, and building chatbots

According to IBM’s Mitch Mason, who makes a lot of them, there are three types of chatbots:

  • Support chatbots; 
  • Skills chatbots; and 
  • Assistant chatbots.

Support chatbots

Built to embrace such tasks as answering FAQs and walking an employee through a process, support chatbots use both short- and long-tail questions-and-answers. The chatbot’s ‘world’ is the organisation and targeted parts of it. As it gets fed more data it can assume responsibility for supporting more and more parts of the organisation. 

Skills chatbots

These are the Siri, Google Assistant and Alexa’s of this world. They can take voice commands and add contextuality. So, by way of example, the pepperoni pizza ordered over our phone in the image at the top of this post, the chatbot knows to have it delivered to the 4th floor, where the human is, not just the front lobby, where the pizza delivery person would probably prefer to leave it. Or in another example, saying, “Turn the lights on” the chatbot knows to turn the kitchen lights on (which is where you are) rather than the living room lights, or else turn the lights on on Level 3, where you currently are, not Level 4 where your desk is. 

Assistant chatbots

These are the most complex of all. Assistants can liaise with other bots to achieve outcomes.

So, for example, and using the office meeting example from an early vidcast of ours, the assistant could greet you by name when you enter the meeting room, send the meeting files to your device (laptop, tablet, or phone, you choose), hand off the translation of the meeting in real time to another bot which forwards the translation to your Indian contractor in Bangalore (and vice versa, handles the translation of your Indian contractor to the meeting), organise the diaries of meeting participants so that post-meeting discussions can take place, and books the meeting room for the next meeting.


No matter what type of chatbot you choose to build (and it will probably be some combination of two of these three), make sure you build in some personality to your chatbot—even though your employees and external stakeholders know that it’s a bot talking, it’s fun to play with it. When something is fun, it’s more likely to be used again later. You don’t have to make it a stand-up comedian, but it should be the personification of the values and ethics of the organisation

Which makes Siri all the more fun to interrogate when you say to it things like, “I see a little silhouetto of a man”

We post updates about once or twice a week, usually at the beginning of the week.