My Role: Product Design Lead
With a simple prompt, AI can make hundreds of choices on a users' behalf. That's why it's so important to craft a UX that gives them the visibility and control to use it with confidence.
Hive.co is a company that allows event promoters to sell tickets through email and SMS campaigns. Their next big step forward would involve leveraging AI to instantly create campaigns that users could send to a highly targeted group of customers likely to buy.
My first step was to work with the leadership team to describe exactly what an AI initiative entailed and why it was so important. Through diagrams and "blue sky" prototypes I helped articulate our vision.
To move forward, the app would need to recommend campaign strategies and instantly generate emails and SMS messages.
The first step to building AI powered campaigns was to figure what choices AI would need to make to craft effective campaigns. This involved mapping out workflows for the engineers to follow as they integrated the AI with the existing app. The merging of a new practice with old software produced many surprises to be ironed out in testing, but we emerged with a functional product and a solid understanding of how it worked.
We were eager to ship what we'd created to beta users, so our initial UX needed to be a light lift.
I took the approach of redesigning the final screens of our existing flows. These screens helped clarify that we had used AI to "fast forward" through the steps users would normally follow, presenting a summary of the choices we'd made.
We began sharing quick campaigns with a beta group and measuring the time it took our users to create and launch campaigns. I reached out and interviewed each Beta customer to gather their reactions and uncover issues.
I already knew there were gaps in the UX and functionality, but the interviews helped me understand how the gaps affected users, and which issues could be addressed quickly and allow them to work faster.
One pitfall of "fast forwarding" through a process was that users lost track of the choices made. I redesigned our loading dialog to summarize what the AI was doing.
Impact: Users stopped backtracking through the process to understand what had happened.
Some of our users represent different acts, venues and festivals, each with their own unique branding. Our generated campaigns needed to match the different templates they had created.
Impact: By supporting templates, we were able to generate campaigns that required less manual editing.
Part of our AI's job was choosing specific recipients for campaigns. We felt the underlying logic was sound, but to ship the bet fast we did not make any changes to the existing UX. This meant when users tried to understand why the AI had chosen certain recipients, they were presented with a long list of filters that was difficult to understand and edit.
The UX for choosing recipients in Hive was a real "power user" experience: a complicated set of filters and conditions that could be leveraged to target very specific groups of recipients. Redesigning it would need to capture, at a glance, why our AI had made its choices, and allow for easy editing.
Our new quick campaigns are available for general usage and we continue to improve on their variety and quality. In less than six months, the vision I proposed when I joined the company has become a reality.