Client: ENDA Technologies
Role & Team: Lead UX Designer (CTO)
Version: 2.0
Timeline: Apr 2025 - Jul 2025, 16 weeks
Result: All users were able to complete key workflows without assistance after redesign leading to 100%+ increase in task completion rate.
During a 4-month contract with ENDA Technologies, I led a redesign of a web app to improve workflows for healthcare practitioners. I made the app more aligned with the doctor, nurse, and medical assistant’s daily practice needs.
40
Accounts
+100%
Trust KPI
+100%
Task Completion
Impact
Redesigning a Healthcare app for Automated AI Note-taking
ENDA Technologies app was really hard to use
The app was so hard to use that customers required training. I adapted the app to better serve healthcare practitioners in their preferred setting—a medical practice.
Recording: Requires doctor to press buttons to start and stop sessions limiting face-to-face interaction with patients and focus on quality care decisions.
Template: Lack of proper customization options for templates leading to excessive manual edits.
Transcription: Long loading times (2-3 minutes) pull attention away from a patient and reduce the number of appointments in a day.
The Challenge
Users must scramble to edit template during an appointment
The user starts and stops the session without the the option to pause the session
The user must wait for the transcription to load in order to proceed
Before Redesign
ENDA Technologies' Original Design → Product Is Rigid To Use
I broke down the client requirements into 3 core flows
I identified inefficiencies in the current app. Then, I demonstrated the main tasks doctors, nurses, and medical assistants needed to complete with the redesigned app.
Doctor: Use ENDA Technologies' app to generate notes for patient and copy and paste into medical records
Nurse: Use ENDA Technologies' app to create a 'Pre-Operative Care' template for patient
Medical Assistant: Use ENDA Technologies' app to look up patient's last consultation notes and correct medication dosage
User Flow Diagrams
Doctor
My next step was to remove as many dependencies as possible
I noticed there were many steps in the app. I mapped out the relationship between different steps on the homepage to ensure proper functionality. In Version 1.0:
Record: Users had to select a template before starting a recording
Resume Recording: Users did not have the option to resume a recording (instead start a new one)
Select Language: Users could not select a language for the template
Dependency Diagrams
Doctor
Before diving into the wireframes, I settled on a scrolling pattern
The scrolling, accordion, and page-by-page patterns were three different ways to organize and present content in the user interface.
Patterns
Pattern 1: Scrolling
Pattern 2: Accordion
Pattern 3: Page-by-page
Content is displayed in a long, continuous view that users scroll vertically
Content is grouped into expandable/collapsible sections
Content is split across multiple pages
After Redesign
Step 1
The nurse can upload a template to be parse in the app.
Step 2
There is a share feature that allows the nurse to copy the link for the template.
After Redesign
The doctor records an appointment by clicking on the 'Start Session' button
The doctor fills in the details based on the conversation with the patient
The doctor can start the process to generate a note by clicking on the 'Generate AI Note' button
The basic premise of the app is for the doctor to record a consultation, fill in consultation details, and generate a note (after selecting a template and a language from the drop down menus). Depending on their needs, the doctor could select multiple templates from the drop down menu. This resulted in automated AI-powered notes to copy and paste into medical records.
Record a consultation
Fill in consultation details
Generate AI note
First, I identified the process for the doctor to generate AI-powered notes
Ideal Scenario: Doctor
The doctor uses a voice command to start the session and end the session for the recording. This allows the doctor to fully use his hands to attend to the patient. Dentists can dictate the recording during the examination without interrupting their work.
1. 71% of patients preferred the doctor did not use the desktop computer (Mobius, 2019)
2. 29% of patients saw no difference when the doctor used the desktop computer (Mobius, 2019)
I proposed hands-free recording with an intelligent assistant
Core Feature: Doctor
Detecting AI Voice Transcription
Start/End Session
The most commonly spoken languages by ENDA Technologies’ customers are English and French. Instead of selecting a language from the dropdown menu, I pitched the option for the doctor to auto-detect the language based on the recording. The app detects cues that the voice transcription is in English. This standout feature allows the doctor to focus his attention on the patient during the consultation.
1. 12.5% are English speakers in Quebec
2. 77.5% are French speakers in Quebec
I allowed the doctor to focus his attention on the patient
Main Screen: Doctor
The option to choose a template from the drop down menu
The app auto-detects English from the recording
Ideal Scenario: Nurse
The nurse creates a 'Pre-Operative Care' template for an upcoming consultation
Based on the doctor’s orders, the nurse creates a new template to generate an AI-powered note for an upcoming consultation. The nurse has the option to parse the data from an existing PDF or paste from another template. Furthermore, the nurse refines the drafted template by simply clicking the contents in the template. The main sections of the ‘Pre-Operative Care’ template are broken down into:
1. Patient Consent
2. Subject
3. Constitutional Symptoms
4. Objective
5. Assessment and Plan
The nurse fills in a prompt in the AI-assisted template builder
The AI-assisted template builder confirms the changes with the nurse
Step 2
The system-generated templates are in the 'Pre-Built Templates' row.
Step 1
The user-generated templates are in the 'My Templates' row.
I offer unparalleled customization options for templates to match patients
Main Screen: Nurse
The nurse can choose from a variety of templates in the template library. The star icon allows her to favorite a template (e.g., Treatment Plan) and move it to the top of the screen. The template library contains user-generated templates and system-generated templates:
User-Generated Templates: Referral Letter, Patient Feedback, Prescription, Treatment Plan
System-Generated Templates: Doctor, Dentist, Veterinarian, Psychologist
The medical assistant searches for a patient with highly accurate search results
Ideal Scenario: Medical Assistant
The medical assistant searches ‘Ale’ in the search bar. The app returns two search results that match the term before she even finishes typing ‘Alex.’ These highly accurate search results ensure that she can quickly find and update past consultations in the app.
Step 1
The medical assistant updated the note for the patient.
Step 2
The app confirms if she wants to proceed with updating the template based on the note.
After Redesign
Main Screen: Medical Assistant
The medical assistant verifies the information from a patient's last consultation
The medical assistant reviews the note from Alex Hirsch’s last consultation. She updates the medication dosage, symptoms, and weather-related conditions in the note. Since she is updating important information in the note, the app detects the four changes and offers to update the template. This highly advanced feature ensures that templates reflect the changing needs of the medical office:
1. Medical knowledge evolves
2. Outdated templates
3. Ensure legal compliance (HIPAA, CMS)
4. Personalization
Research
After testing with ENDA users, I needed to adjust the templates and show the AI-powered voice transcription
Based on two-part concept/usability testing with 9 participants, I learned that users preferred to see a preview of the template in the template library. This allowed them to understand the nature of the contents before opening the template. Furthermore, I learned that users wanted to see the note and transcription side by side for optimal editing.
1. Template previews
2. Side by side AI-powered note and voice transcription (with timestamps)
I changed the cards to include a preview of the template before opening
Before Testing
After Testing
Core Feature: Nurse
The nurse relies on AI-assisted template builder to add a new section to the template
As she is refining the template, she starts to run out of ideas. This is an optimal opportunity for her to rely on the AI-assisted template builder on the bottom of the screen. She enters a specific prompt, responds to a follow-up question, and confirms the changes in the template. The AI-assisted template builder uses context (user input or data) to dynamically generate tailored sections for templates:
1. Input Collection: User provides data
2. Context Understanding: AI parses the data to extract key information
3. Section Development: AI builds a section for the template
4. Post-Processing: AI improves the coherence, tone, and style of the section
5. Output Delivery: Final section is presented to the user with editable text area
Template Library
The user views their favorite template along with other templates stored in the app.
Search
The user searches for a patient by searching for the first three letters of the title of the consultation.
Record
The user records an appointment by clicking the 'Start Session' button.
After Testing
Responsive Mobile Screens
I adapted the content on 3 screens to fit the mobile phone for ENDA Technologies' users
ENDA Technologies’ customers struggled to view the app on their mobile phones. The app content did not adjust to fit on the mobile phone screens. I designed 3 screens by shrinking the features to adapt to the mobile phone platform, ensuring the app was responsive. I also tested the 3 screens in the Figma app on my phone.
1. Record
2. Template Library
3. Search
Design System
I created a new design system to provide consistent components for building the product
I realized that I had designed many components and variants throughout the project. I needed to create a design system to house all these components for the developers.
The Result
Impact
+100%
Task Completion
+100%
Trust KPI
40
Accounts
Version 2.0 offers smarter functionality than the previous version of the app. The new features powered by improved algorithms (e.g., search, template previews, live captions) allow the app to perform decisions that are more intuitive and tailored to the healthcare practitioner’s daily practice needs.
Search Term
Bolded Search Results