Conversational UI: Create Successful Chatbot Experiences

Conversation design is critical to creating an effective chatbot. After a bad experience, 30% of consumers tell the company and 50% tell their friends (Qualtrics, 2017). Designing the chatbot with a user-centred approach creates more meaningful interactions with customers and minimises the risk of a poor customer experience. This article outlines the 7 key considerations to conversation design to begin your chatbot journey or take your current chatbot to the next level.

  1. Identify chatbot goals

Decide on what the chatbot is trying to achieve. Collect historical FAQs to find out what pain points the chatbot will solve, and the anticipated results. This step also involves deciding what form the chatbot will take: an autonomous chatbot (all machine) or an assisted chatbot (human and machine). Autonomous chatbots can be risky if the user is asking complicated questions the bot can’t answer, so make sure you choose the form with the user’s experience in mind.

  1. User research and personas

Identify who will be using the chatbot and create a persona around them. Use contact history to find out the general characteristics of target users and what they tend to ask/need. Finding out their context (the information surrounding their questions) and how they speak (e.g. shorthand, slang, acronyms) and using this to influence the design will create a more customised, user-centred result, producing an overall better user experience.

  1. Topic mapping

Map which requests the chatbot will understand and which answers it will be able to provide. Content for responses can be existing (e.g. brand history or policies), adapted (existing content structurally modified to fit a conversational flow), or created (e.g. personality and humour). Group phrases and meaning to better align intent with actions, with at least 10 versions of a question for each intent.

  1. Design the dialogue

The bot should speak the same language as the target users and its personality (tone) should reflect the brand. Consider the purpose of the dialogue: does it inform users or take actions? Also consider its proactivity: will your AI assistant direct users or let the user guide the experience?

  1. Design the user flow

Decide on the result of users interacting with the chatbot to lead them in the right direction. Start by informing users what the chatbot can do for them, then guide the user through the conversation. For example, suggest topics to start the conversation or offer actions to choose from through multiple choice questions. It’s important to not overload the user with questions as this would create an overwhelming flow. Instead, ask the user to fill out a form to create a more simplified experience. End the conversation with a definitive goodbye or transferring to agent.

  1. Design the responses

Keep answers concise and ask the user if they would like more information rather than putting it all into the one response. Inject personality into the chatbot to make the interaction feel more natural. Manage failure by being honest with the user and apologise for not understanding, provide suggestions on what it can ask, and offer the ability to contact a human.

  1. Monitor and manage

Monitor customer usage patterns to provide insight into how to update and improve your chatbot. Build in continuous learning processes to ensure that your virtual assistant gains capabilities as your business develops.

When designing your chatbot, these terms may be found useful to describe scenarios.

  • Utterances: Terms a user would type.
  • Intent: What is the user asking?
  • Fundamental Meaning: Grouping of multiple intents to provide an overall reason the user has typed what they typed e.g. “I would like to request a refund on my bike purchase because I asked for a repair ticket but could not get a response in the time I wanted.” – Multiple intents, but fundamental meaning is “User wants a refund.”
  • Dialogue Flow: Workflow of a conversation. Based on the utterance, what is the intent and what is the task that the user wants the agent to do?
  • Entity: Nouns or objects in a sentence. Useful for putting structure into an unstructured conversation.
  • Traits – Topics or background of a conversation.
  • Channels: How your UI is deployed i.e. Messenger, WhatsApp, Web etc.

We hope this information inspires you, wherever you’re at on your chatbot journey!

You can watch the live presentation of this topic below.

Oct 26, 2020

Author: Kate Edwards, UX Designer,