Zoom Virtual Agent - Form Builder
Creating forms to be used with the chatbot required engineers to custom code form variants for each client. This lead to hours of work for the engineering team weekly, and weeks of waiting for clients to use custom forms for their users within the bot.
Overview
Zoom Virtual Agent is a multi-modal bot product that uses AI in order to help clients provide service for their customers, resulting in saved money for the company. When the bot is unable to help and support is needed from a human, the bot will offer various support channels that the end user can choose from.
Most third party CRM integrations (Salesforce, Kustomer, Zendesk, etc) require basic information be sent from the bot to the CRM so that the agent providing support has basic details about the customer and doesn’t have to ask again for all of the information they just provided to the bot.
The Problem
Basic forms were hardcoded to send the abosolute minimum to CRM’s on the handoff, such as the customer’s name and issue type. This worked initially, but once we started adding more CRM’s with different requirements, this was challenging to customize per client and was hard to scale. Customers also expressed that they wish they could send more data from the bot, to make the handoff more seamless and increase the service level agreement per engagement.
Discovery Research
I started by syncing with the engineering lead for integrations on what the perceived issue/opportunity was. He let me know that engineers spend many hours customizing forms for every customer and support channel, and that this was not a scalable approach. After finding out about the different CRM’s that we support and how the integrations are set up, I was ready to talk to some customers about this.
I talked to 5 different customers about their feedback on the support integration between the bot and their CRM.
Discovery Findings
Findings from the discovery research yielded some common themes.
They wanted different support channels to have different forms for more specific handoffs.
The current form experience felt rigid and had to have many custom fields hardcoded.
The turnaround for requesting a new form took weeks.
Engineers spent hours weekly working on these form customizations.
Design Solution
After syncing with the team on the findings from the research, we brainstormed how this might work. After a few different design explorations we decided to go with a custom form builder, which solves for all of the pain points above along with future proofing it for the future.
After bot builders integrate to their CRM initially, the available fields are prepopulated into our form builder for them to place.
Fields can be conditionally set so that based on a response, the rest of the fields would update - leading to less required fields.
Forms can be published by the admin with no coding or PSO help necessary.
Live UI preview.
Universal for all CRM’s.
Design Validation
After building an initial wireframe, I pitched the idea to our internal team for feedback. I got some great feedback that we ended up using - such as the ability to autorefresh the integration in the background so the admin doesn’t have to think about it.
We then demoed this to the initial three clients and they had some slight critiques, but overall were ecstatic with the amount of time this would save them, along with how powerful it will be.
“Wow! I had no idea you guys were working on this. This will be a game changer for us and probably others you work with. ”
UX Metrics
How did I know I was successful or not once the project was released?
Aside from the qualitative follow up interviews, tracking overall adoption and usage of the new feature seemed logical.
We tracked if new requests were being opened for our engineers to custom build new forms.
We tracked how many forms each customer created per bot.
We tracked how often they were updating forms based on new fields being available in the CRM.
Results
There were a few discoverability tweaks we needed to make for the new form builder - as it was missed by a couple of customers, and new form submissions were being requested. After updating that, and guiding customers to the new form builder, it was met with a lot of optimism and joy as customers now could build their own forms for their bot.