Reducing navigation stress and wait times at the University of Michigan Hospital through an accessible, multilingual AI chatbot solution.
Aspects of this project are confidential & can’t be displayed publicly. If you’d like to know more about my process, schedule a portfolio presentation.
The hospital ran a volunteer program where undergraduate students stood at specific locations throughout the facility to help visitors navigate. While helpful, this approach was limited by volunteer availability and couldn't serve all patients, especially non-English speakers.
We wanted to integrate with the hospital's IT infrastructure so that users have an easier way of accessing navigation features, but that was not allowed as we were an outside organization and that would directly violate HIPAA.
We wanted to use using Bluetooth beacons for real-time positioning, but our limited funding did not allow that solution.
We wanted to automatically pull patient appointment data and proactively provides directions, however we were not allowed to access user data as that would violate HIPAA.
Why it works: no personal data required so it's HIPAA compliant, simple infrastructure, quick to implement, easy for users to adopt (no app download, just scan and go), low cost, is multilingual by design to help non-English speakers.
Hold context from conversation
Have close to 0 response time
Always provide option to talk to human/get out of the chatbot flow
Warn before repeating an answer
Saying "I don't understand" is less frustrating than giving the wrong answer
Politely tell user to only ask for direction if they ask any other information
An AI-powered chatbot that provides instant navigation assistance through simple QR code access. No app downloads, no account creation, no personal information required.
Scan QR code placed throughout the hospital
Ask where you want to go in your preferred language
Get instant directions to your destination
Multilingual support serves non-English speakers in their preferred language
HIPAA Compliant, no personal information collection or storage
Zero friction access - no app downloads or account creation required
Would require significant data collection effort beyond our resources
Mapping every room would be time-intensive and difficult to maintain
We would ask the user to enter the number on the poster to identify their location, but that means added friction (extra user action), potential for language barriers and user error (entering wrong number)
Different QR codes at different locations automatically provide the chatbot with user's current position
How it works: Reception QR code tells chatbot user is at reception, South Entrance QR code indicates South Entrance, etc. Users get contextual directions without any additional input.
Limited access to users taught me to leverage secondary research strategically and validate assumptions through creative methods like Wizard of Oz testing once the MVP proved valuable.
We considered complex features (history tracking, appointment integration), but the HIPAA-compliant standalone approach proved most effective. The best solution isn't always the most feature-rich.
The location-specific QR code insight came from observing real user behavior. Mid-journey navigation failures revealed that users don't need smarter AI, they need the system to already know where they are.
As students with limited resources, we focused on solving one problem exceptionally well rather than tackling multiple problems poorly. A 50% reduction in navigation time proved we didn't need hospital data access to create meaningful change.
