Creating a chatbot isn't just about designing something; it's also about designing someone. That's why startup x.ai hired Harvard grad Anna Kelsey, whose education centered around folklore mythology and theater. Kelsey's experience with characters and how they react and respond to one another helps her
The other side of the chatbot design coin is one that UX designers should be more familiar with. By mapping user flows, UX designers help chatbots provide suggestions and prompts to better understand and address users' needs.
Still, mapping dialogue isn't the same as stringing together
Even for experienced product designers, building a personable chatbot is a big undertaking. Take these four steps to design one that your customers truly want to talk to:
1.Start with the most important question.
Think long and hard about your chatbot's first prompt. Remember that it's the foundation for every interaction that follows.
If you’re a retailer and anyone can visit your site, you might use a general question like “What can I help you with today?” Customers could be browsing your selection, inquiring about a return policy, or looking for the nearest store. A multiline insurance agency that’s built a chatbot for its customers might start with a different line, such as “Hello! What type of insurance do you have with us?”
Your question should
2.Think about the user’s agenda.
The chatbot we built for San Francisco's procurement department, PAIGE, was successful not merely because of the AI technology behind it, but also because of the time and effort we put into its workflows. As a result, PAIGE recognizes “buckets” of questions — about commodity purchasing, for instance — and then walks the user through the process with a decision tree.
We obviously weren’t new to the idea of workflows, but constructing PAIGE underscored to us the importance of robust workflows in a technology product. Any conversation has two agendas, one for each speaker. The chatbot's agenda must be built around the user's expected goal or need. Building a series of questions that will guide users to their desired outcome is vital.
3. Consider your brand’s voice.
What does your brand sound like? If you’re running a chain of tattoo parlors, it might be irreverent and rebellious. A university chatbot could be scholarly with lots of school spirit. Using the right language when writing your copy is key to drawing users into the experience. Users should feel like they're interacting with your organization, not a robot.
Coca-Cola has been around for well over a century, but in that time, it's only had one real voice. The Coca-Cola brand is designed to evoke happiness, whether that's spending time with friends or enjoying the holiday season. From messaging on billboards to advertisements featuring happy, Coke-drinking polar bears, the brand has stayed true to its voice throughout the years. A Coca-Cola chatbot would need to carry this brand voice by evoking these themes as well.
4. Be conversational.
Interacting with a chatbot shouldn't feel like reading a textbook. Think about how the
If you follow Taco Bell on Twitter, you've seen how it uses creative and funny retweets, delivers clever comebacks to insults and jabs, and even offers life advice. In spite of the fact that Taco Bell is a corporate giant with 7,000 restaurants and almost $2 billion in revenue, its Twitter feels decidedly less corporate — which is why users interact with it.
Whatever you have in mind for your chatbot, focus on personality and user flows. Do that, and you'll build a bot that's exactly who your customers are looking for.
Tony Scherba is a President | Founding Partner at Yeti. Tony has been developing software since high school and has worked on digital products for global brands such as Google, MIT, Qualcomm, Hershey’s, Britney Spears and Harmon/Kardon. Tony’s writing about innovation and technology has been featured in Forbes, Huffington Post and Inc. At Yeti, Tony works on strategy, product design and day to day operations, hopping in and working with the development teams when needed. Follow Tony on Twitter.