How to approach the design

AI-generated Key Takeaways

• Begin conversation design by building simple automation like an FAQ bot and iterating based on testing and refinement.

• Enhance the agent with rich features for more interactive journeys, eventually integrating identity verification for a fully contained and personalized experience.

• Define the agent's use case, design its personality, craft a welcome message, and map out the ideal conversation flow.

• Use simple language, provide clear options with suggested replies and actions, and prepare fallback responses for unexpected user input.

• Continuously test the agent with real users, track relevant metrics, and use feedback to refine and advance the conversational experience.

For a strategic approach to conversation design, start simple and iterate. Start by designing basic automation, such as an FAQ bot that answers common questions from a predefined list. Once you've tested and refined a basic agent, you can build on it. Use rich features to add a more interactive journey, such as a product search. At this stage, the agent may send users to your brand website to complete a task. Eventually, you can use OAuth to perform identity verification, enabling the whole journey to happen within the agent for a more personalized and convenient experience.

What follows is an overview of conversation design for Business Messages. It describes key components of the iterative design process and a general order of operations.

Build the basics

• Define the use case and related tasks the agent will support. Prioritize a use case with potential for large impact.
• Design a personality for the chatbot that reflects your brand and feels appropriate for the use case.
• Craft a welcome message that sets the user's expectations. Clearly state what the agent can do, and provide conversation starters for tasks the agent supports.
• Map out the ideal path the user will follow from start to end of the journey. To get a feel for the conversation's flow, write some sample dialogs first.
• Write short messages to advance the user step-by-step along the conversation path. Use simple language that mimics the rhythms of natural human speech.
• Use suggested replies and actions that provide clear options for what the user can do next.
• Expect the user to type unexpected things. Prepare fallback responses for any freely typed messages, or "freetype," that the agent doesn't recognize or support.
• Create a handoff strategy for scenarios that require a live agent. After two fallback responses, surface a live agent request suggestion.
• Test the agent with real users.
• Track metrics related to conversational quality. Learn what you need to improve by reviewing survey data and mining users' freetyped messages for insights about the agent's performance.
• Use the feedback to refine the conversational experience.

Advance

• Add another use case to the conversation flow.
• Create a visually dynamic experience using rich cards. Use carousels of rich cards that let users browse a range of options.
• Test the agent and iterate on the design.

Excel

• Add a more complex workflow to the use case.
• Integrate with OAuth to authenticate the user's identity, so they can access personal account information securely within the conversation.