Cvent Assistant

Cvent was looking for ways to enhance product experience with the introduction of AI and chatbots in some key product areas. This project aimed to set the foundation for how to design chatbots and conversation design more broadly, and support designers in creating mocks for stakeholder presentations. This then became the blueprint for UX’s AI chatbot pattern library.

The problem

With increased interest in providing AI experiences, we needed a way to provide our product designers with a ready-made pattern from which to build chatbot interfaces. Up until this point, individual product teams were building the interfaces and experiences independent of each other resulting in a variety of solutions and approaches that lacked continuity across the product experience and Cvent branding.

The approach

We can’t design for the actual chatbot conversation, especially when we use LLMs instead of a direct script; however, we still want to be able to design the concept with a plausible conversation between a chatbot and its user. I wanted to provide a template with just enough conversation elements that covered as many flows as possible to give product and content designers a starting point for their designs.¹ To achieve this, I needed to understand the goals for the chatbot, and the primary problem areas we wanted to solve with the chatbot in Attendee Hub.²

Workshops

I facilitated two workshops, and used Figjam to brainstorm and organize information.

Workshop 1: Product Managers, Client Services, and Sales

Objective: Understand the business value and needs of a chatbot interface in the product Understand how we talk about about Attendee Hub Understand what kind of help attendee users typically seek. We discussed information categories that should be prioritized within the chatbot, and which categories were considered secondary and tertiary to the chatbot’s purpose. This allowed us to map the different conversation flows3 and to prioritize the chatbot’s primary function.

Workshop 2: UX Designers

Objective: Explore how information would be presented in a chatbot interface. We also discussed information categories but looked at it from an experiential perspective, namely, what’s the best format to present certain types of information? How can we appropriately build engagement opportunities? How can we best support our users without introducing cruft and unnecessary interactions that might make the chatbot unusable?

WORKSHOP SUMMARY

Design: Conversation

To create the script flow, I used the template provided by UXCC in its Conversation Design for Chatbots course¹. Before jumping into drafting the mock conversation flows, I did preliminary work to establish scope and set expectations for what this chatbot should do once built and deployed.

In steps 1-3, we determined which elements to include in our chatbot, and mapped out the primary actions, secondary actions, and catch-all and error handling actions. We also outlined the voice and tone of the chatbot. Doing this before designing the conversation allowed us to stay on track and stay aligned with the group.

I then created tabs for as many known flows as we might have–though, because this was meant to be a template, it is not a total representation of a “complete” script flow. We just needed enough for designers to use to create plausible prototypes to present to stakeholders without going through the laborious effort of drafting a script each and every time they wanted to design a chatbot experience.

Design: Chatbot interface

To tie it all together, I designed a chatbot prototype using our Design System library. I explored key elements, including button iterations and how best to present certain types of information (e.g., FAQs, session details, maps, etc.).

Final output

Chatbot pattern library

The design iterations became the basis for our AI pattern library, which has been widely adopted and used by designers across multiple products.

Chatbot script template

The script I designed for this specific project became a mock template. Teams started using bits and pieces of the script as a way to populate their prototypes rather than trying to come up with their own elaborate conversation. This helped them get their concepts to stakeholders faster to facilitate discussion and decisionmaking for the final product.

Lessons learned

Applying what I learned in the UX Content Collective workshop was an object lesson. Sometimes, when you’re given a template at a workshop, it’s difficult to assess how and when to use that template in your own work. The template itself is relatively simple and easy to use but I found its application initially difficult. Once I figured out how to organize the conversation flows and had an idea of what pathways to pursue, the template became a lot easier to use–and revealed an important aspect of the project at large. It’s easy for us to go straight into designing the interfaces but this project showed very quickly how untenable that approach was. Drafting out the script and, at minimum, 50% of the pathways was critical for helping visualize the UI for the chatbot. Without the script itself, I’m not sure we would’ve been able to design the chatbot interface pattern as quickly as we did. It also made it easier to discuss how to present certain types of information within the chatbot window.