I discovered chatbots in 2016 and immediately fell in love with them. I was immediately interested in understanding the UX implications they would have and how we could, as product managers, define the ultimate characteristics of a chatbot that looked and sounded empathic. In this post I suggest a few design tips and tricks for chatbots that feel less like a machine.
- First rule: Using short text blocks makes my both feel more human. Long blocks of text == boring!
- If your chatbot is going to send several blocks of text make sure to use the typing indicator as a way to both indicate to the user more info is coming their way and in order to give them more time to read.
- Early on I had to design a personality for my chatbot. This helps me keep the tone uniform across topics and defines the tone of my links and default answers (for example when something is not understood)
- Because my chatbot is for young women I decided to make my bot a chatter and a joker. She also uses loads of emoji! 🤖❤️Adapt this to the personality of your audience. I recommend you make a very deep analysis of how your users chat when you do your user personas.
- At certain points I need to make sure the chatbot understands the question asked. For this the easier is to repeat the understanding and to offer yes and no buttons. When doing this, to get rid of the phone tree feeling, I make my bot say things like “Did I hear xyz?” or “That sounded like xyz – have I gotten that right?”
- Avoiding repetition in hard as hell, but specially important for confirmations and high level menus. While testing the bot I realised some sentences are repeated again and again – which makes my chatbot a bit boring. Write variations of frequent answers.
- Last, but probably MOST IMPORTANTLY: Language variations are going to represent a big challenge. Spanish, the language I’m using for my chatbot, varies a lot from one country to another and even within some countries, as in Spain. In my case, as a Spanish from Spain writing a chatbot for Uruguay, this is particularly challenging. The only solution for this, for now, is basically to work with locals and end users, understand their use of language and iterate, iterate, iterate. Maybe in the future AI will take care of that.
Overall remember that the golden rule to conversation design is features, navigation, small talk, as these are the three elements, in that specific order, you need for a delightful experience.
If this is a topic you find interesting, I presented a talk about challenges when developing NLP chatbots you can watch here:
Did you enjoy this content? Reach out at firstname.lastname@example.org if you’d like to have me at your conference or event. Let’s check my agenda and plan a talk that your audience enjoys!