Publishing AI, Made Simple
in watsonx.ai
Turning a 16-step deployment process into a single click for developers.
OVERVIEW
It shouldn't take longer to publish an AI application than it does to build one.
Developers using watsonx.ai's visual tools, like the Prompt Lab, faced a frustratingly complex process when trying to publish their applications as API endpoints (live web addresses other software can connect to and use) so they could work with them directly from their own computer or coding setup.
I led the optimization of this 16-step process, enabling users to publish their generative AI applications all in “one” click.
ROLE
Product Designer
UX Designer
TIMELINE
3 months, 2024
SKILLS
UX design
Developer experience
Workflow optimization
TEAM
2 researchers
4 developers
CHALLENGE
Deploying an application from Prompt Lab was unclear from the very first step. Instead of staying within the visual tool, users were pushed into a scattered workflow across multiple browser tabs: reading documentation in one, editing and running code files in another.
Several of these steps required a level of technical knowledge users often didn't have, and nothing in the interface signaled what came next, or whether it was even necessary.
How can developers easily deploy applications built inside Prompt Lab as API endpoints?
To address this, I focused on simplifying the flow to:
Reduce time to deploy
Minimize confusion