The five most common conversation patterns, and how to use them
Planning is paramount when creating actions in Watson Assistant. Use these common conversation patterns to design exceptional user experiences.
Conversations that go somewhere
Watson Assistant’s mission is delivering outcomes without friction. With Actions, the design process is now squarely focused on creating a successful, delightful conversation between your assistant and your end users. Content and conversations, not coding and configuration, are the center and circumference of the build interface, meaning your development team’s creative bandwidth is squarely focused on the user experience.
A successful, delightful conversation between users and your virtual assistant depends on careful planning and considered design. Just as you planned your assistant before beginning the build, you need to plan your conversations before diving into creating them in the actions editor. And just as we designed the Assistant’s UI with you in mind, you want to design conversations with your users’ goals as your North Star. When you and your team have done that prep work and are ready to start building an action in Watson Assistant, you’ll have an easy, smooth build if you align the conversation to one of our five most common conversational patterns.
Learning these patterns will equip you with solid, streamlined templates for building actions that satisfy your users’ most common requests. We’ll be covering the patterns in order from least-to-most complex:
1. Basic question and conditional response
2. Information gathering with conditional response
3. Information gathering with warm agent handoff
4. Step-by-step instructions
5. Recommendations and troubleshooting
We’ll showcase the structure of each pattern through hypothetical examples based on the many industries where Watson Assistant is successfully delivering exceptional customer experiences at the enterprise scale. Before we get into that, here are three key principles to guide you when designing your conversations:
Be concise: you have limited space in your panel, your users have limited time, and your assistant’s conversation needs to move quickly.
No FAQs: clarification is the name of the game in actions. Questions that only require a single response are best handled by the Search skill.
Always be testing: the preview panel is automatically configured with all the new content you train the assistant on in the actions editor. Use it to ensure that your conversation is flowing logically.
Got it? Great. Now, let’s see the top five conversational patterns in action.
Basic question and conditional response
Watson’s ability to apply conditional logic allows users to get concise answers to questions with more precision than in an FAQ. In the following example, an employee has questions regarding their benefits (employee-facing HR assistants are a popular use of Watson Assistant across industries).
Information gathering with conditional response
This pattern showcases Watson Assistant’s ability to steer the conversation by offering clarifying options in order to provide users with the most accurate information possible. Just like the first pattern, a simple user inquiry starts the action, but this time the assistant prompts the user with options that will help steer the conversation to its appropriate conclusion. Our use case is a healthcare provider’s customer-facing assistant.
Information gathering with warm agent handoff
Watson Assistant is more than an engine for driving users’ experiences: it’s also a powerful ally to human agents and customer service representatives. In this pattern, the assistant gathers information from the user before handing the conversation off to a human agent.
In the following example, Watson Assistant serves as a buffer at the beginning of the conversation, saving the agent time and tedium by obtaining account information before handing off the conversation to a mobile/telecom service rep. The rep can finalize the transaction without spending time retrieving the customer’s information.
Step-by-step instructions
This pattern takes the burden off of human agents completely. Watson walks the user through every step of the conversation, asking clarifying questions along the way and obtaining confirmation from the user before proceeding. This pattern is ideal for use cases involving financial transactions such as when the user wants to either initiate or dispute a transaction. In the following example, the assistant guides the user through the process of formally disputing a transaction.
Recommendations and troubleshooting
This pattern puts Watson Assistant’s full range of skills on full display. In the following example, the assistant gathers information and asks clarifying questions in order to zero in on an insurance recommendation for a user seeking coverage. Note how the assistant’s responses are contingent on the information provided in previous answers, and how, when the user provides an invalid response, the assistant offers alternative options to help steer the conversation back on track.
Conversations built on trust
Again, Watson Assistant’s mission is to eliminate friction. By centering the building process on content rather than coding and configuration, Actions allows you to create an experience focused from beginning to end on the user’s desired outcome. Applying these patterns when creating your assistant’s actions will help build trust in your users that the assistant can solve even their most complex requests.
At the top we reiterated the importance of planning when building and publishing your content. Let’s also remember that a healthy lifecycle requires you to continuously improve your assistant’s actions. Getting to launch is just the end of the beginning. Watson’s story takes place in five acts:
- Watson is trained
- Watson is tested
- Watson is launched
- Watson gets results
- Watson keeps learning
Building conversations centered on your users will help them trust your assistant, and continuously improving your assistant will deepen that trust. If you’re new to Watson Assistant and looking for inspiration, explore our clients’ success stories for examples of how some of the world’s premiere companies have leveraged Watson Assistant to improve their customer- and employee-facing solutions. If you’re ready to start building or have just gotten underway, refer to our previous articles ‘Plan It’ and ‘The Build Guide’ for a refresher on best practices for getting your proof of concept up and running.
