Timeline
May - August 2026
Role
Design Manager (leading AI style guide creation and co-managing designers & stakeholders)
Collaboration
1 PM, 8 Designers, 1 Lead Engineer, 2 Product stakeholders, Leadership team
Problem
Individuals experiencing housing instability have a hard time navigating the house search and application by themselves, especially when human support is limited.
Solution
An AI assistant that explains housing processes and requirements, provides personalized guidance, and closes the communication gap between the renters and housing case managers.
Estimated Impact
25%
of the case managers' workload is expected to be automated
60%
estimated AI adoption rate
PROBLEM
Support Provided <<< Support Needed
At Housing Connector, case managers are the middleman between renters and property owners, providing support and connections.
On average, 1 Case Manager support 20 Renters at the same time.
Can AI fill the gap by supporting and guiding renters through the house search/application processes when case managers aren’t available?
Based on survey responses from 30+ renters and interviews with 10 case managers…
Renters expressed no concern about using AI for housing support.
Case managers believed AI would be a great additional source of assistance.
AI GUIDELINES & BRAINSTORMING
AI is for basic requests and manual processes only
Although renters and case managers are open to using AI in the house search process, human intervention is still preferred when it's available.
As this is the first-ever AI assistant on the platform, we defined rules and boundaries for how it should (and should not) function.
AI can…
Explain/simplify housing terms
Handle monotonous tasks and basic questions
Recommend relevant resources
Redirect renters to human staff
AI can't…
Predict the chance of approval or rejection
Communicate with case managers & property owners without renter’s involvement
Make decisions on behalf of users
Provide solutions to complex, situational problems
As a co-manager of the project, I led the creation process for the AI components and style guide
DESIGN DECISION #1
AI entry points that get users familiar with AI and actively offer real-time support
Most renters come from a low tech background and are not familiar with AI, so we were considering the most accessible position to place the AI assistant.
Although the app has an onboarding flow to introduce the chat interface to users, actually using the AI chatbot (and remembering it's available) is a different story.
I created some AI entry points placed them across the app where it makes sense:
Offer real-time support in specific use cases, instead of letting users go ask for support themselves (they may not know they can)
Introduce AI to users throughout the user flow, not only in onboarding
Potentially increase AI adoption
DESIGN DECISION #2
Email Draft closing communication gap between renters and case/property managers
A problem that case managers kept stressing is that a lot of renters and property owners are not responsive to emails and messages, and there's a lack of consistent communication.
A designer on the team and I brainstormed ways AI could help close this gap.
— Case manager A
Iteration 1
Give users the contact info of case/property manager, and users reach out by themselves
"Not enough autonomy"
Iteration 2
Draft and send emails on behalf of renters, using bot email addresses
Stakeholder’s feedback: Automated emails between chatbots doesn’t mean effective communication between humans.
"Too much autonomy"
Final Iteration
AI drafts message and redirects users to Gmail app. Renters still take initiative to reach out.
"Just enough autonomy"
Email draft deserves its own component instead of being nested in a message.
Push notifications (in-app, email, text) to nudge users to reply to messages and stay engaged in the house search process.
IMPACT
Lighter workload for Case Managers and better support for Renters
After testing the design with 12 users and presenting it to stakeholders across the nonprofit, we got some promising results on the performance of the AI assistant.
of the case managers' workload is expected to be automated
users found the AI assistant helpful and would use it in their house search
THE PEOPLE <3
To those who made this happen with me
It was my first time leading a team of 8 designers, and I learned a lot about design and people management. Thank you to the team members who made this project successful, and the stakeholders who trusted us to work on the product's first-ever AI features!
