Defining interaction patterns for conversational UIs

I led the team dedicated to designing chatbot experiences at a large insurance company. The systems used a variety of technologies, including natural language processing. Unlike simple marketing bots, they were designed to helped users with their request no matter how they described the issue—meeting their need if possible and transferring to a support rep if not.

how i helped

Leading a team dedicated to this work, I quickly realized that at a base level, we needed a shared design language. Most conversational interfaces are flows, so I created a shared library of components that team members could put together for designs.

Once were had a shared language for design, along with guidelines for how to use individual elements, my next step was to define interaction patterns that teams could adapt and user over and over again.

This patten for intent recognition helps users recognize what the system understood, giving them visibility and control.

This patten for intent recognition helps users recognize what the system understood, giving them visibility and control.

I began to explore patterns based on two main criteria:

  • Value to the user

  • Usefulness to our teams

The challenge is that products and visions were different across the company, so we had to ensure a pattern was high-level enough to be useful across teams, yet detailed enough that it retained its value.

The results

My work enabled stakeholders and teams to have a shared language for designing chat interactions. Re-usable patterns helped teams leverage what was working for users in one area when designing for another, greatly increasing their efficiency and effectiveness.