Insider, Artificial Intelligence

5 best practice design principles for building chatbots that work

16 Aug, 2017
6 Min Read Dan Ward

5 best practice design principles for building chatbots that work


Recent Posts

Subscribe to our blog

Imagine that you’re eating lunch in the kitchen at work. Suzanne from HR walks past and comments that your food looks delicious, sparking a natural and easy conversation about the best cafes in the area.

Moments later, you’re interrupted by another co-worker, Gary. He wants to know if you’d like to grab a bite to eat. You explain that you’re already eating lunch, and he changes the topic to his five cats. The conversation is stilted and uncomfortable.

If you’re like most people, chances are you’d prefer interacting with Suzanne. When conversations are easy, relevant and enjoyable, we’re more likely to want to participate.

It’s the same for chatbots. Annoying, unhelpful and awkward bots can do more harm than good. That’s why it’s essential to follow best practice chatbot design principles from day one.

Here are our top tips for building a conversational user interface that enables chatbots - and in turn, your business - to thrive in the brave new world of artificial intelligence.

1. Define clear entry and exit points

For a seamless user experience, a chatbot needs to operate within clear boundaries. What is its role and what kind of problems does it solve? When will it become available, and how will the user know? When will the chatbot complete its interaction with the users?

When defining a chatbot’s rules of engagement, focus on how it navigates entry and exit points. For example, does the chatbot automatically pop-up when a user enters a website? What mechanism does it use to notify users that it is available? How will it know that a user no longer needs assistance?

There’s no single answer for how to configure entry and exit points, as long as they’re clear, consistent and predictable. Be conscious of creating a positive user experience and avoid overwhelming users with interruptions and pop-ups.

2. Don’t underestimate the importance or complexity of natural language

Chatbots win over users when they deliver a human-like interaction, so strong language comprehension abilities are non-negotiables. This doesn’t just mean creating chatbots with large vocabularies. Rather, the chatbot needs to be able to deal with the quirks of human language.

Teach your chatbot to recognise common typos and mistakes to enable smooth, fluid conversation – even if the user makes a transcription error. Get the chatbot familiar with internet slang (i.e. ‘omg’, ‘lol’, ‘that’s lit fam’) and make sure it can comprehend and mirror specific speech patterns and phrases of users on its platform.

Remember that language continually evolves, so review and refresh your chatbot’s vocabulary often.

3. Create a chatbot with spark and personality

Users expect chatbots to provide a human-like experience, so it’s natural for them to expect that the chatbot will have human-like traits. What could be more human than the rich expression of personality?

Chatbots with a unique, distinctive and engaging character are more appealing to users. They help to build trust and rapport, facilitate the experience of a genuine interaction and simply make the conversation more enjoyable.

If you want your chatbot to capture your target audience, assess your company’s brand and cultivate a personality for your chatbot that supports it.

4. Be upfront about the limitations

Artificial intelligence has made leaps and bounds, but the fact is that your chatbot will still have some limitations. There will be occasions when the chatbot can’t comprehend a request, or when it makes a mistake in translating a user’s intentions. Rather than ignoring or overlooking these limitations, tackle them head on.

Make sure your chatbot is upfront about what it can handle (and equally importantly, what it can’t). Create implicit boundaries by asking guided questions (e.g. “Can I help you with A, B or C?”) to encourage users to stay within the chatbot’s limitations.

If the chatbot doesn’t understand or can’t complete a request, be transparent (i.e. “Sorry, I don’t understand.”) and provide the user with alternative options to ensure their needs can be met (“Would you like to start again?” or “Please contact customer service on the following phone number”).

5. Continuously improve and learn from your mistakes

The best way to ensure your chatbot succeeds is to evaluate and improve its interactions. This can be done overtly (i.e. seeking feedback from users) or covertly (i.e. analysing points of failure where the chatbot can’t meet a user’s needs).

Take an active role in monitoring and measuring performance and implement specific, targeted improvements to address them. Feedback from users will let you know if you’re on the right track.

What next?

As the popularity of chatbots increases and adoption by mainstream businesses grows, developers and programmers should expect consumers to have higher expectations about the usability and capability of chatbots. There’s a premium on getting conversational user interface right, and the time to act is now. RXP can help you build chatbots that are both easy to use and support natural conversations.

For more information about how to deploy a chatbot in your organisation, Digital Platform Trends 2017: Diving into chatbot-driven conversational commerce