How do you design a bot personality that can work in multiple channels?
If the conversation you write for your bot is your script, the personality you add is the scenery. The personality of your bot should reflect your brand image, and make your bot more engaging to interact with. However, functionality must be prioritised above everything else — so can the same personality work in every situation?
I faced this challenge as we considered rolling out our chatbot onto different channels. Each had differing levels of functionality, and just as I would have considered how a website would look and act on different devices, I wanted to consider how a chatbot should ‘look and act’ on different channels.
I identified 4 different Conversational User Interfaces that I wanted to explore, each with increasing level of visual interactivity.
→ Voice Only
→ Text Only
→ Simple UI — buttons or other very simple UI elements
→ Rich UI — image carousels, actionable UI elements (payments, etc)
I started with an idea from web development, progressive enhancement. My initial thoughts were that voice-first would be the same in Conversational UX as mobile-first is in Web UX. Interestingly, while this idea did map somewhat to the content, the personality element needed its own tuning unique to each situation that wasn’t a straightforward progression. As such, taking the idea of responsive design was a much closer fit.
In a voice-only interaction, brevity and clarity are much more important than personality. You should of course keep your script on-brand, but functionality needs to always come first.
Adding additional ‘chit-chat’ may initially seem like adding to the conversational tone, but it can easily be counterproductive.
Ever got a voicemail where at the end you have to go and get a piece of paper and a pen, and re-listen to try to get down all the details before you forgot them? And then had to listen again, because even then you couldn’t get them all down on the first re-listening? Yea? Don’t do that.
Users care much more about getting the answer they need than the social niceties that normally accompany human conversation.
It’s also worth considering what actual physical voice the channel you have chosen will be using, and how what you have written will sound in that voice. It may not fit with the same tone you have used elsewhere, and so adjustments may need to be made.
It may not seem obvious, but this is where personality is the most important, and can be the most built out. Since you only have text, the actual conversation is what needs to keep the user engaged and continuing to interact with your bot. Although your responses still need to be easily digestible, the user is reading rather than listening, so you can add more text without significantly reducing the ease of picking out the answer.
Personality becomes especially key when the conversation goes wrong. Without the fixed input of buttons or cards, the user can say anything, and natural language understanding, while improving all the time, still has a way to go! Being able to gracefully let the user know you either didn’t understand, or that they have gone “off script” is critical. As is gently leading them back into the conversation your bot is designed to handle. A little bit of humour when things go wrong can do a lot to ease user frustration.
If you are using the UI elements sporadically here, then you can keep the personality very similar to the text-only channels. However if you are using the buttons almost exclusively to drive the conversation, rather than letting the user answer in natural language, there will be adjustments needed. It may not be obvious, but when you are restricting the user’s responses, the exchange is less conversational. So although you will want to keep the same tone, you may wish to dial back the personality as it isn’t needed as much.
A mistake I often see made is using your bot’s own ‘quirky’ personality in response option buttons. Please don’t force your personality on your user. The buttons you provide are your user’s response to you — if you are making them use your voice that can be a pretty uncomfortable user experience.
This is the bot equivalent to the person who gives you a nickname 5 mins after meeting you. Seriously, we’re not friends. Don’t be that guy.
If you are offering the user a yes/no choice, then buttons with ‘Yes’ and ‘No’ are all you need, and the quickest and easiest for your user to process. To the question “Do you want to continue” I might be ok with choosing “yes”, but feel less good about having to choose “I can’t imagine anything I would rather do”. It would also take me precious moments longer to work out which answer correlated to ‘yes’ .
So keep the personality on your side of the conversation, and be aware if your bot personality is particularly “strong” you may need to tone it down to keep the conversation balanced.
Here personality can take a backseat. Very much like your voice interface, the actual ‘conversation’ should ideally be kept brief and to the point. Pictures will always draw the eye quicker than text. Ideally your Rich UI should do a lot of the talking. The user may not even really pay attention to anything other than the text in the cards once they are presented.
Use your rich UI elements to reflect your personality and tone, and you won’t need as much conversational script overall.
Just like a good human, a good bot should be ‘socially aware’ – adjusting for different contexts. For the very best Conversational User Experience you are going to have to do some work tailoring your content and personality to each situation — much as it would be great to be able to use the same conversations everywhere without any work on your side!
Have fun with your bot personality, but don’t get carried away. Make sure your bot is primarily fulfilling the user need, and functionality that it is designed to.
Source: Chatbots Magazine