1 Best Practices Conversational UX Design 2 Best Practices - PowerPoint PPT Presentation

1

Best Practices Conversational UX Design 2

Best Practices Conversational UX Design SET THE RIGHT EXPECTATIONS Set the right tone Tell users what they can do and what to expect Is it more an open or a closed conversation? 3

Best Practices Conversational UX Design What is the context of your user? Where is the chatbot integrated? CONSIDER THE CONTEXT 4 UX Bear: http://uxchat.me/ Image Context Matters: https://chatbotsmagazine.com/context-is-king-creating-intelligent-conversations-for-chatbots-9372bd6c2ca4

Best Practices Conversational UX Design MAKE SENSE OF USE CASES Image: https://www.gizmodo.com.au/2016/04/facebook-chatbots-are-frustrating-and-useless/ Does the new flow make sense? What use cases the bot can do and can’t do Is it more efficient than using other channels? 5

Best Practices Conversational UX Design PERSONALITY IS CRUICIAL What personality should your chatbot reflect? Image: https://magazine.chatkit.com/2018/11/28/how-to-give-your-chatbot-a-personality/ Create a persona that best fits your product 6

Best Practices Conversational UX Design DESIGN LANGUAGE Take the time to design language Find the right tone of language UX Bear: http://uxchat.me/ 7

Best Practices Conversational UX Design EFFICIENT COMMUNICATION Do you want to order pizza or pasta? Clearly set the expectations for your bot Formulate clear questions and answers Yes. 8

Best Practices Conversational UX Design TALK ABOUT FAILURE Image: https://chatbotsmagazine.com/how-to-write-user-friendly-error-messages-41e66a77a026 A user’s question always needs to be answered Acknowledge failure in a remorseful tone 9

Best Practices Conversational UX Design SMOOTH HANDOVER Define clear handshakes between human and bot Integrate your bot in your customer facing strategy Image: https://www.kommunicate.io/blog/chatbot-human-handoff/ 10

Best Practices Conversational AI - Training & Testing 11

Best Practices Conversational AI Training & Testing Current “ natural language understanding” ≠ true language understanding Conversational AI calculate a user’s intention taking into consideration the user’s input • • the intent it is taught to expect the training it receives based on intent • 12

Best Practices Conversational AI Training & Testing Intents Information Action Booking a Flight Checking Flight prices flight in details 13

Best Practices Conversational AI Training & Testing Intention: flight prices Calculated intent: flight details Intention: flight prices Calculated intent: checking in Intention: flight prices Calculated intent: flight prices 14

Best Practices Conversational AI Training & Testing Variety of natural language makes a broad language training extremely difficult The more varied language inputs are, the better the application will be able to recognize intent Finding large groups of different people to interact with is difficult in prototype and beta phases Chatbot & Virtual Assistant Testing • uncover training gaps provide invaluable training data • 15

Best Practices Conversational AI Training & Testing Chatbot of a retail chain Intents Opening hours Locations Special offers Careers 125 individual 25 individual 25 individual 125 individual questions questions questions questions 16

Best Practices Conversational AI Training & Testing 36 testers came up with more than 3500 questions regarding store locations Overly creative questions Where are stores in 48.158538, 11.645419? • I‘m not allowed to vary • Is it possibe to use the bus from Bad Homburg to go the same question with shopping in one of your stores? different locations? What stores are near the highway? • I don‘t think there are 125 different ways to ask for a store location. 17

Best Practices Conversational AI Training & Testing Build your bot up slowly, but consistently • Start with the core intents Train all intents sufficiently to avoid under- and overtraining • Use as many different users as possible • Limit their individual input ; train in iterations if possible • • Consider what kind of training fits your intents best Does the intent encompass isolated information? • Does the intent trigger a process? • Never stop training your bot • 18

10 Do’s & Dont’s when designing a chatbot 19

1 CHOOSE THE RIGHT PLATFORM: Right platform, right technology – choose wisely 2 CHOOSE THE RIGHT PEOPLE: Several disciplines needed – technology, screenwriting, linguistics 3 SET THE RIGHT EXPECTATIONS: Users need to know clearly what to and what not to expect 4 PERSONALITY IS CRUCIAL: Start with a bot persona - don't let your bot pretend to be a human? 5 DESIGN LANGUAGE: Target-group & chatbot personality specific language design 6 TALK ABOUT FAILURE: Compensate flaws and error loops – give clear feedback to the users 7 SMOOTH HANDOVER: Integrate in customer facing strategy, define handshakes with human 8 MAKE SENSE OF USE CASES: Design clear use cases where usage of the bot make sense 9 CRAWL, WALK, RUN: Build a core MVP with intents, build layer by layer 10 TEST, TEST, TEST: Test in iterations and with your specific target group 20

We are here to help Testing Reality: Real Users. Real Devices. Real time Chatbot & Virtual Assistant Testing Contact us for more details or live demo Our all-inclusive language test for your chatbot, Do you have questions about Chatbot and virtual assistant conversational AI or language assistant: Use our crowd to testing? Would you like to have more information on how test and optimise the capabilities of your application under we can help you make your digital products even real conditions! better? Just send us a message. Discover more > Get in touch > 21