AI note editor redesign
Mentalyc is an AI clinical documentation platform for therapists and mental healthcare professionals. The product helps clinicians create AI-powered clinical notes, treatment plans, therapeutic alliance insights, and supervision documentation.
I served as Lead Product Designer, owning the end-to-end design strategy for the note editor experience. I led research, solution definition, usability testing, and cross-functional collaboration with the product manager and engineering team to redesign the note viewing and editing interface. A second phase added Magic Edit, an AI tool for the fixes clinicians kept making by hand.
Year
2025
Timeline
1 month
Tools
Figma,Mixpanel,Clarity
My role
Lead Product Designer,UX/UI Designer,Researcher,Design System Designer
Team
Product Manager,Developers
Redesigned note editor — before VS after
Problem statement
Design challenges
- Very limited time and focus on quick delivery
- Focus on “low-hanging fruits” improvements
Business
problems
The market shifted when major EHR (Electronic Health Record) systems began embedding AI documentation tools directly into their platforms. This created financial and workflow friction for Mentalyc users who now questioned paying for two separate tools.
While Mentalyc had expanded beyond core note-taking with advanced features like Treatment Plans, Alliance Genie (therapeutic alliance insights), and Supervision tools,
the product faced critical retention and adoption challenges:
- Churn was rising
- Not sufficient or low adoption of “advanced” features (Treatment Plans, Alliance Genie, Supervision)
- Users churning specifically because they didn't
perceive sufficient value and couldn’t justify paying extra for the note-taking tool - Low awareness of note customization capabilities,
leading to dissatisfaction with note quality (AHA moment)
Business
goals
Reduce churn
within 3 months post-launch
Increase in-app note editing
(vs. editing in EHR)
Increase template customization adoption
Reduce support tickets related
to note editing confusion
Improve discoverability of features accessible
from note editor (Analytics, Treatment Plans, etc.)
Increase user engagement with note-related settings
Research
Research process
I proposed a research-first approach
to understand how clinicians actually used the note editor and what barriers prevented effective editing workflows.
CS ticket analysis
100+ support messages over 6 months. Top issues: templates and note types (50), client reassignment (34), note content (10).
UX audit
Heuristic analyses of the product and flows.
Competitor & desk research
Established mental models for discoverability and navigation.
Clarity heatmaps
Identified low-engagement zones and banner blindness.
User interviews (4)
1:1 sessions with active clinicians, walking through their own notes on a shared screen
Survey (~250 responses)
“Help us shape Magic Edit”: how clinicians edit, which edits they repeat, and what they’d want AI to do. The most common answers became its suggested prompts.
User testing (12)
Two rounds of 6: the new editor, then Magic Edit
Key open questions
How users’ workflow looks like? How do users currently approach writing and editing notes in Mentalyc?
Do users understand that notes are editable within Mentalyc?
Can users discover key features in the left toolbar (Transcript, Analytics, Treatment Plans)?
What's the mental model for where to find settings
(global vs. local)?Do users understand the connection between notes
and templates?What are the most common edits users need to make?
Research insights
Discoverability barriers
Note editability
2 of 4 didn’t know they could edit notes in Mentalyc,
and 3 of 4 edited in their EHR instead.Left toolbar features
Icon-only design hid the tools: 2 of 4 struggled to find the transcript, and 1 copied it by hand, not knowing it could be downloaded.
Templates
1 of 4 didn’t know templates could be customised,
and nothing showed which template a note came from.
Global settings
None of the 4 would go to the global note or data settings (terminology, modalities, data preferences) on their own.
Note header
2 of 4 didn’t know they could customise what the note header shows.
Saving notes
Only about 60% of users saved notes to records. Those who did treated it like a to-do list, not knowing why it mattered.
Highlight
Important observations
- Some users viewed their EHR as source of truth, not Mentalyc
- Users approached Mentalyc with "copy-paste" mindset rather than exploring features
- Poor information architecture caused users to look for options in wrong places
- The same small fixes came up note after note: placeholder names, wrong pronouns, mixed-up speakers, notes that ran too long
Solution & testing
Design goals
- Make note editability obvious through clear visual affordances (editable text styling, formatting toolbar, UI copy)
- Improve feature discoverability in the left toolbar by adding labels to icons and organizing content clearly
- Surface the template connection — show which template was used and provide easy access to template editing
- Create entry points to global settings directly from the note editor — terminology, modalities, and data preferences
- Reduce repetitive manual editing by introducing AI-powered Magic Edit — find & replace, pronoun changes, adding and removing details
- Improve information architecture by reorganising session details, client details, and note settings into logical, accessible sections
- Clarify note header customisation so users understand what information can be shown and how to configure it
Magic Edit
Shipped later in June 2026
Clinicians fixed the same things in every note by hand. Magic Edit lets them describe a change in plain words, or start from a suggested prompt, and review it before it touches the note.
- Suggested prompts come from the survey’s most common answers: find & replace, make more concise, include missing details, change pronouns, anonymize
- A before/after preview highlights every change; the note only updates once it’s saved
- Past versions stay one toggle away, so nothing is lost
Usability tests
Participants & Method
- Two rounds of 6 participants: the new editor, then Magic Edit
- Moderated task-based usability testing using the clickable design prototypes (online)
Results - note editor
Worked: Users reacted positively to updated UI, describing
it as "cleaner and more organized"Worked: 5/6 users quickly spotted they could switch between Clinical/Narrative/For Client note views (significant improvement)
Worked: Users reacted positively to new labeled toolbar layout, noticing options they'd never seen before
Worked: Finding session transcript caused no hesitation (major improvement)
Concern: Some confusion about whether settings apply locally
or globallyConcern: Mixed preference on placement of download/copy actions
Concern: Some users associated template editability with
"Edit Note" button causing slight confusion
Results - Magic Edit
Worked: All 6 users understood how to accept/discard edits individually or in bulk
Worked: 4 users understood the process fully
Worked: Common labels like "Make concise" and "Change pronouns" were clear
Concern: 2 users could not locate Magic Edit without guidance
Concern: All 6 users found some suggested edit labels
confusing or risky
Changes after testing
Note editor
Changed: Moved the formatting toolbar above the note, where its actions are easier to spot
Changed: Replaced the "Edit note" button, which users read as a step before editing, with a "Change template or client" link under the title
Changed: Cut the formatting options down to the ones users said they’d use: text style, bold and lists
Changed: Showed more client and session details in the header by default
Magic Edit
Changed: Rewrote vague prompts: "Improve writing" became "Refine writing & language", and "Remove details" became "Remove sensitive information"
Changed: Merged overlapping prompts into one: "Include missing details"
Changed: Replaced bracket placeholders, which users over-deleted, with an open prompt box
Changed: Every edit previews before and after, with past versions one toggle away, so the original note is never lost
Final designs
Designs
Achievements
Key result
Monthly churn rate
−10%
Before launch (Jun–Aug 2025) vs. 4 months after (Dec 2025)
Breakdown per metric
Before launch (Jun–Aug 2025) vs. 4 months after (Dec 2025)
Note editing
Template customization
Global note settings
Magic Edit
- AI edits accepted
77%
- Active clinicians using it monthly
12%
Jul–Aug 2026
