Improving AI Usability for UM’s Community
Mi-Maizey
Overview
Summary
Improving the Usability of the “Create a GPT” Feature
Mi-Maizey is a generative AI platform developed to assist University of Michigan faculty, staff, and students in managing projects and navigating campus life. In collaboration with University of Michigan Technology Services, our team focused on redesigning the "Create New Project" feature, which was reported as unintuitive and difficult to use. Over a four-week sprint-based Agile project, we enhanced usability, accessibility, and user alignment through research-driven design improvements.
Research
The Problem
UM-Maizey “Create a GPT” Feature is Not Intuitive
The current "Create New Project" feature on the Maizey platform presents several challenges that impact its usability and effectiveness. Users may encounter difficulties due to unclear form fields, limited guidance, and a lack of streamlined workflows, which lead to inefficiencies and errors during the project creation process. Additionally, the feature does not fully align with the University of Michigan's brand guidelines, impacting its visual consistency and professional appeal. These issues highlight the need for a redesign to improve the feature’s usability, clarity, and overall alignment with user needs and expectations.
How Might We
Simplify the Process
Usability Testing
Users Found the Layout and Verbiage to be too Complex
Participants: 5 University of Michigan students from technical backgrounds (School of Information & Engineering)
Methods:
Three key tasks: project setup, layout configuration, AI customization
• Collected both quantitative data (task duration, success rates, error counts) and qualitative feedback (user quotes)
Key Findings
AI Configuration Complexity: 0/5 participants were able to configure AI settings without errors.
Difficulty Locating Key Features: Only 1/5 participants successfully found the “Add New Data Source” button on their first attempt.
Need for Clearer Instructions: Participants struggled to understand AI prompt augmentation settings, requiring external documentation for clarification.
Desire for Examples: Users suggested adding sample use cases to demonstrate how AI configurations work.
Designs
Decision #1
Streamlined Form Structure
Reordered fields logically, starting with critical decisions (e.g., "Public or Private Project?")
Reduced on-screen information to minimize cognitive load
Decision #3
Simplified AI Configuration Experience
Simplified AI settings language using tooltips and step-by-step instructions
Added examples of AI capabilities for clearer user understanding
Decision #2
UI Enhancements for Better Usability
Incorporated University of Michigan branding for consistency and trust
Organized information into human-readable formats
Decision #4
New “Link to Database” Feature
Incorporated University of Michigan branding for consistency and trust
Organized information into human-readable formats
Decision #5
Summary Page
Incorporated University of Michigan branding for consistency and trust
Organized information into human-readable formats
Impact
Although implementation is ongoing, our redesign provides the foundation for:
• Improved Usability – Simplified workflows and better guidance reduce confusion.
• Increased Accessibility – Redesigned interfaces make MiMaizey easier to use for diverse users.
• Brand Alignment – Consistent visual design fosters trust and aligns with University branding.
Reflection
Lessons Learned
Agile sprints allow for iterative improvements but require prioritization.
Clear, user-centered language is critical for accessibility.
Brand and accessibility integration from the start leads to more cohesive designs.
Challenges
Small usability test group (5 participants) limited insight scope
Participants were primarily from technical backgrounds, lacking broader representation
short timeframe
Next Steps
Conduct usability testing on the redesigned prototype with a broader user base.
Perform a heuristic evaluation to validate improvements.