Targeting non-technical and budget-conscious Azure users, I designed an AI agent that automated 70% of user's cost optimization process, and guided them to make smarter, more cost-effective decisions.
Timeline
June - September 2025
Role
UX Design Lead, UX Research
Collaboration
Azure Growth Experience team: Designers, Researchers, PM, Engineers
Located on the Microsoft Azure portal, it's a tool that helps users track, manage, and optimize their spending on cloud services.
It offers insights to help users stay within budget and reduce unnecessary costs.
While anyone can be a user of Azure, those who experience the most problems are:
Students & individual users
Startups & small businesses
Compared to other Azure users, they have…
Limited budget 💵
They are extremely mindful of their spending.
Limited experience 💻
They need guidance and support with cloud cost management.
Limited resources ⏳
They prioritize speed, autonomy, and simplicity.
users are interested in AI-based cost optimization suggestions.
(A startup developer)
How Might We…
The design process in a nutshell
iterations from sketches to high-fidelity
design critiques
8
usability testing sessions with diverse users
~25
sources for reference
20+
stakeholders & collaborators
3
new solutions contributed
"Push the boundaries and go beyond the existing patterns. Just be creative."
This is what my mentor told me on our first meeting. I was the first designer on the team to focus on building AI features for Cost Management specifically. Given the creative freedom, I explored how AI could be integrated naturally across the site and exactly where users need it, instead of making them turn to the chatbot for help.
Cost Optimization User Flow
Cloud users who are mindful of their spending need clear visibility into pricing whenever purchasing or modifying services.
With Cost Simulator, users can enter an action they plan to take on the portal and visualize potential cost changes before moving forward.
What else can Copilot Agent do beside generating cost simulation?
What makes it different than regular chatbots?
Beside passively answering user's questions, the AI Agent in Azure Cost Management can autonomously suggest ways to optimize costs whenever possible, and directly make changes to cloud services with the user’s approval.
This feature is especially helpful for users who are new to Azure, lack cloud experience, or simply don’t have the time to manage cost optimization themselves.
The core priority of the process was to help users build trust with AI and feel more comfortable using AI features by maintaining transparency and human control.
Users have the ability to…
Preview the process AI will be taking before starting the task
Monitor, stop, or take control from AI in real time
Chat with AI to clarify confusion and provide feedback
Another design problem with the Cost Management portal is the distribution and organization of content.
Most pages on Cost Management
❌ Packed with a massive amount of data and information
❌ Highly technical and complex content
The current homepage
❌ Fails to capture the most important, relevant information to users
❌ Provides generic and introductory resources like tutorials, which are mostly for first-time users
Newly designed Homepage
I tested the prototype with 8 users from diverse backgrounds—including students and tech professionals—with varying levels of cloud experience (beginner to intermediate) and from different regions across America, Europe, and Asia. The outcome was promising.
87%
Participants said the user flow was easy to comprehend, and the visual design and content were intuitive
70-75%
of the user process was automated by Copilot Agent, which gave users ease of mind
At the same time, I also received some valuable constructive feedback that could help me and the team improve the design further:
The transition between tabs and subpages under Cost Overview could be confusing
Some features/ buttons were vague or not noticeable
Some text content is lengthy
Responsible AI
Being assigned to work on an AI-first project, I always reminded myself the most important goal was to remain user-centric and make sure my AI-powered design could solve user problems.
Think outside the box
As a designer, I’m often encouraged to follow the design systems, but in this project, I was given the creative freedom to explore new design elements. This experience taught me to take bold steps beyond the patterns I had limited myself to, create something new, and demonstrate its usability to the team.
Two-way collaboration
As a young designer, I’ve always seen collaboration as a way to learn from others and get feedback on my work. But this project showed me that my ideas could also contribute to others’ projects and inspire them. My design turned out to be more influential than I expected!
Although this was my second internship with Microsoft, the experience didn't remain the same. I worked on a different product space, had a new manager and some new team members, and even got to connect with some of them in an in-person summer party! I learned a lot from this half-new-half-familiar internship experience.
