At a glance

Role
Product Designer

Team
1 UX Designer, 1 Design Lead, 3 Developers, 2 UX Researchers, 3 Writers, 1 PM, 1 SME

Timeline
3 months, 2024

Skills
UX/UI design, product design, workflow optimization, developer experience, wireframing

Challenge

Developers using watsonx.ai’s visual tools, like Prompt Lab, faced a frustratingly complex process when trying to deploy their work as consumable API endpoints.

Solution

I led the design of a one-click deployment experience for watsonx.ai, enabling users to publish generative AI services without navigating complex workflows.

I collaborated with cross-functional teams to streamline a 16-step deployment process into “one” click.

impact

💥 65% reduction in time to deploy

💥 81% reduction in steps to deploy 

💥 Designed a reusable pattern used across multiple watsonx.ai visual tools

 
 

background

Problem statement

How can developers easily deploy services built in Prompt Lab as consumable APl (Application Programming Interface) endpoints? 

Business Opportunity

By simplifying the flow and keeping users in context, we aimed to:

  • Reduce time to deploy

  • Minimize confusion

The current experience

Before the redesign, users had to initiate deployment through a non-intuitive “Save as” function, select a notebook type, create a deployment space in a new browser tab, and complete 11+ more steps, without in-product guidance. 

In total, this process involved 16 steps, multiple tools, and assumed extended Python expertise, which created confusion, drop-off, and a poor developer experience.

Design explorations

I analyzed the current user experience to identify where users were getting stuck and which steps caused the most confusion. Based on those insights, I explored two distinct design directions.

idea 1

Integrating the deployment space creation flow into the Prompt Lab via a tearsheet triggered by a “Deploy” button.

  • I aimed to keep users in context by avoiding browser tab navigation, but this design only added visual complexity. 

  • Ultimately, this idea didn’t meaningfully decrease the time to access the consumable API endpoint.

  • Idea 1’s influence on the final solution: I switched to a lightweight version of the full deployment creation flow.

idea 2

Routing users to the “Save As” or “Deploy As” section in a tearsheet based on button selection in the UI header. 

  • I aimed to combine actions into a single component, but there was no contextual guidance nor was it clear if saving was required before deploying.

  • Ultimately, this approach increased cognitive load and raised more questions than it answered.

  • Idea 2’s influence on final solution: I decoupled the deployment and saving user flows into distinct experiences.

These early explorations revealed that even when the deployment process was fully contained within the tool, they still didn’t reduce the number of steps, the time to deploy, or the mental effort required by the user.

Solution

I clearly defined the new 3-step user journey:

 

In total, I :

  • Led the shaping of the redesigned deployment workflow by focusing on user experience, system automation, and simplifying the end-to-end flow.

  • Proposed a design iteration that removed unnecessary navigation across tabs, reduced cognitive load, and assumed no prior knowledge of Python notebooks or deployment infrastructure. 

  • Worked closely with the product manager and developers to align on a more intuitive user journey and identify which steps could be automated on the backend. 

  • Designed the confirmation step, where users receive immediate feedback and can test the API endpoint directly after deployment.

Reflection 💡

What did I learn?

Share ideas with others sooner - I realized it’s better to share ideas early, even if I didn’t think they were “ready”. Waiting until an idea feels semi-polished can lead to wasted time and energy, especially if it ends up going in a completely different direction.

Talk with your developers - When I was first assigned this project, I struggled to brainstorm improvements because I was unfamiliar with the deployment process. Speaking with the developers who built the original experience helped immensely; they were able to walk me through the backend, which not only clarified the technical side of the deployment process but also allowed me to translate that understanding into human-centered designs.