As the lead designer, drove the vision and interaction strategy for the Model Picker within the Azure AI Foundry Agent Builder—a core configuration tool enabling users to select and customize the AI model powering their agent.
Situation
Users struggled to identify the most suitable AI model for their agents due to unclear options, lack of comparison tools, and overwhelming customization settings. This led to confusion, suboptimal configurations, and frequent support requests.
Task
We aimed to address user pain points around complex model selection by making the process clear, guided, and flexible. Business needs focused on reducing setup time, decreasing support inquiries, and improving user engagement with advanced AI configurations. My approach was to build an experience that bridges technical depth with usability. Success would be measured through improved completion rates, user feedback, and reduced onboarding friction.
GOAL 1
Reduce user confusion and errors during model selection by 30%.
GOAL 2
Shorten the average model configuration time by 20%.
GOAL 3
Improve user confidence in their AI model choices, as measured by post-onboarding surveys.
ROLE
Lead UX Designer
TOOLS
Figma, sketches
TEAM
1 engineer, 1 UX researcher
TIMELINE
4 weeks – June 2025
Action 1 – Competitive research
Action 2 – Conclusions and decision framework
Decision #1 – Choose a form factor: Dropdown menu, modal dialog or use prompt settings
Decision #2 – How to show model details: Hover menu, tabs or dedicated pane
Decision #3 – What to show for the details: Use scenarios, filtering, grouping, recommendations
Decision #4 – Is the ability to compare models important in this context?
Action 3 – My recommendations and sketches
- Use a dedicated pane for the model details
- Pick 3 model detail patterns and test them
- Experiment with using tabs for model details
- Experiment with compact parameter at the top
- Experiment with Auto-select and recommending model
- Skip model comparison for now: Only 2 companies, Langchain and GitHub, provide this feature
Action 4 – User research
User research to confirm and identify key decision factors for users choosing a model
(Key factors were: use cases, performance, cost, latency)
Action 5 – Medium fidelity sketches and feedback
Result
Delivered a cohesive Model Picker experience that improved discoverability, reduced configuration errors, and established a scalable framework for future model integrations. The competitive analysis informed design direction and positioned Agent Builder with a best-in-class selection interface.
I designed a guided, comparative model picker that highlights key differences, offers contextual help, and streamlines customization, enabling confident and informed model selection.
