Thread Medical | Clinical Note Management

Crafting note editing workflows with clarity and clinical trust for the AI medical scribe tool

Thread Medical | Clinical Note Management

Crafting note editing workflows with clarity and clinical trust for the AI medical scribe tool

PLATFORM

Healthcare dashboard

Healthcare dashboard

TIMELINE

May 2022 - Aug 2022

May 2022 - Aug 2022

TEAM

Founders, 1 Product director, and 1 Engineer

Founders, 1 Product director, and 1 Engineer

Thread Medical is an AI-powered medical scribe tool that transcribes doctor–patient conversations into structured clinical notes. As the product approached Beta, two critical gaps stood in the way of real-world adoption: physicians couldn't edit generated notes, and there was no reliable way to manage them.

Thread Medical is an AI-powered medical scribe tool that transcribes doctor–patient conversations into structured clinical notes. As the product approached Beta, two critical gaps stood in the way of real-world adoption: physicians couldn't edit generated notes, and there was no reliable way to manage them.

Thread Medical is an AI-powered medical scribe tool that transcribes doctor–patient conversations into structured clinical notes. As the product approached Beta, two critical gaps stood in the way of real-world adoption: physicians couldn't edit generated notes, and there was no reliable way to manage them.

I led the UX work end-to-end. That involved running user interviews and competitor analysis to understand clinical workflows, aligning with the founding team on scope, then designing the note editing experience and overhauling note management. Usability testing with physicians returned an early satisfaction rate above 85%, which validated our core assumptions and helped build stakeholder buy-in heading into Beta.

I led the UX work end-to-end. That involved running user interviews and competitor analysis to understand clinical workflows, aligning with the founding team on scope, then designing the note editing experience and overhauling note management. Usability testing with physicians returned an early satisfaction rate above 85%, which validated our core assumptions and helped build stakeholder buy-in heading into Beta.

Challenges & Opportunities

Challenges & Opportunities

Challenges & Opportunities

Beyond basic functionalities — refining for clinical fit

Beyond basic functionalities — refining for clinical fit

While Thread Medical’s MVP could record and transcribe conversations, it fell short in supporting how physicians actually review, modify, and manage notes in practice. The existing inbox was a simple chronological list, and there was no way to edit notes—limiting its usability in real clinical settings where accuracy and flexibility are crucial.

This gap presented both a challenge and an opportunity: how might we evolve the product from a passive transcription tool into an active part of a physician’s workflow? By mapping out the note lifecycle across different phases of a patient encounter and studying how clinicians interact with existing tools, our goal was to identify key moments where better structure, control, and clarity could drive adoption and trust.

Current production: Note inbox listing notes chronologically

Current production: Note inbox listing notes chronologically

Current production: Recording and transcribing conversations into note

Current production: Recording and transcribing conversations into note

BEFORE THE ENCOUNTER

Prepare

View patient history or previous notes.

Scan for key issues or follow-ups and identify areas to address during patient visit.

DURING THE ENCOUNTER

Capture

Start recording at the beginning of the visit.

Engage with the patient while Thread Medical transcribes in real-time.

AFTER THE ENCOUNTER

Review & Finalize

View the AI-generated clinical note.

Copy the note into the EHR system.

Make edits to match clinical standards or billing requirements.

Discover

Discover

Discover

Understanding the problem space

Understanding the problem space

To understand barriers to market fit and clarify the needs of target users, I synthesized past research and conducted a preliminary competitor analysis during the first two weeks.

While awaiting pilot user interviews, I collaborated with the team to compile a research plan and draft a guideline for conducting in-depth interviews to make sure the research goals are clear with our limited timeframe, while also seeking resources to better understand how AI conversation recognition works.

Key pain points and feature gaps identified through user interviews

Key pain points and feature gaps identified through user interviews

After presenting the interview insights to the team, we identified two critical pain points that, if addressed, would have the greatest impact on the overall product experience. The RICE scoring model helped us to prioritize the user needs considering the limited timeframe and engineering resources at hand.

After presenting the interview insights to the team, we identified two critical pain points that, if addressed, would have the greatest impact on the overall product experience. The RICE scoring model helped us to prioritize the user needs considering the limited timeframe and engineering resources at hand.

Primary

📝 Lack of editability causes trust issues

📝 Lack of editability causes trust issues

The inability to edit AI-generated notes creates friction:

  • Users are uncomfortable signing off on autogenerated notes they can't correct.

  • There’s a strong need for customization, including note formatting and reusable templates tailored to clinic or personal preferences.

Primary

📑 Users need efficient way to manage all notes

📑 Users need efficient way to manage all notes

Current tools lack an efficient note inbox for quickly locating and managing notes, which slows down workflows in busy clinical settings. Users need:

  • A summary view of patient info before visits to prepare efficiently.

  • A quick way to create, locate, and revisit notes.

  • The ability to revisit and update notes over multiple days.

⛓️‍💥 Non-standard note format slows adpotion

⛓️‍💥 Non-standard note format slows adpotion

Users are already familiar with their current EHRs. When the generated notes don’t follow expected structures:

  • It increases mental load.

  • Users are slower to adopt the tool, or revert to manual documentation.

🔌 Lack of integration slows down workflow

🔌 Lack of integration slows down workflow

The current product operates in isolation from existing systems:

  • Users want seamless integration with their EHRs to support sign-off and billing processes.

How Might We support physicians in creating, editing and managing notes with greater efficiency and accuracy?

How Might We support physicians in creating, editing and managing notes with greater efficiency and accuracy?

Focus Area 1

Focus Area 1

Focus Area 1

Note inbox: Making patient notes more accessible

Note inbox: Making patient notes more accessible

Based on insights gathered from physician interviews and past research, we identified a core job physicians try to accomplish when managing clinical notes:

Current production: Note inbox

Jobs to be done - Note inbox

When managing a high volume of clinical notes, I want to easily locate pending or specific notes, so that I can prioritize my work and reduce time spent searching.

Outcome Expectations:

  • Quickly locate notes based on patient or appointment info

  • Identify and prioritize pending or incomplete notes

  • Reduce time spent navigating and switching between tools

Common rules:

  • [Search by Identifiers] Lookup by patient name, DOB, MRN, or visit date

  • [Note Status Indicators] Clear visual cues for pending, and completed notes

  • [Quick Access] Minimal steps to open, review, and take action on a note

Redesign: Note inbox

Focus Area 2

Focus Area 2

Focus Area 2

Creating a note editing experience that feels instinctive to physicians

Creating a note editing experience that feels instinctive to physicians

As part of the core clinic experience, the note editing flow needed to balance clarity with control. Our research showed that physicians primarily review AI-generated notes and only occasionally make edits—lacking a clear distinction between viewing and editing could potentially lead to confusion and accidental changes.

I explored several design directions to address this, ultimately landing on a flow that separates passive review from intentional editing—reducing errors and aligning better with real-world physician workflows.

Current production: Note page

Jobs to be done - Note page

After a patient visit, I want to review and edit the generated notes to ensure accuracy and completeness, so that I can sign-off and move on to the next task.

Outcome Expectations:

  • Distinct separation between viewing and editing to prevent unintentional changes

  • Maintain alignment with physicians' familiar documentation workflows

  • Improve confidence and accuracy in final notes before sign-off

Common rules:

  • [Clinical Context] Access to referenced information (e.g., medications, history)

  • [Edit Mode] Dedicated "Edit Note" CTA to prevent accidental edits

  • [Interaction Patterns] Familiar controls that support quick additions and corrections

From note inbox to individual notes

Focus Area 2

Focus Area 2

Focus Area 2

Diving deeper: handling diverse input fields

Diving deeper: handling diverse input fields

After nailing down the high-level editing flow, I explored a range of interaction patterns to make note editing feel more intuitive and aligned with physicians’ mental models. A major focus was the Review of Systems (ROS) section, where dense, structured content often leads to inefficient documentation.

During early usability testing with physicians, I noticed that my initial design for editing ROS wasn’t as intuitive as expected. Users hesitated, made errors, or had to ask clarifying questions—all signs of friction. This feedback made it clear that the flow didn’t reflect how physicians naturally edit their notes.

After nailing down the high-level editing flow, I explored a range of interaction patterns to make note editing feel more intuitive and aligned with physicians’ mental models. A major focus was the Review of Systems (ROS) section, where dense, structured content often leads to inefficient documentation.

During early usability testing with physicians, I noticed that my initial design for editing ROS wasn’t as intuitive as expected. Users hesitated, made errors, or had to ask clarifying questions—all signs of friction. This feedback made it clear that the flow didn’t reflect how physicians naturally edit their notes.

Exploration 1 🔴

✅ Immediate editing of each symptom directly in the list

✅ Visible groupings by system

❌ Why it wasn’t adopted:

Low efficiency: Requires users to interact with every individual symptom one by one

Hard to quickly add new symptoms: "Add symptom" per section adds friction

Error-prone UI: Easy to mis-click a radio button while reviewing rather than intentionally editing

Exploration 2 🔴

✅ Clear separation between “reports” and “denies” per system

✅ Easier to edit or remove individual symptoms quickly

❌ Why it wasn’t adopted:

Too much reliance on dropdowns: Manually selecting system, symptom, and status every time is slow and repetitive

Hard to quickly add new symptoms: "Add symptom" per section adds friction

Limited discoverability: Symptoms aren’t surfaced unless selected—reducing awareness of available options

Final design 🔵

Batch editing at scale: Allows users to select multiple symptoms at once, and apply "Reports" or "Denies" in bulk

Reduced friction: Centralized modal supports focused, efficient symptom entry

Familiar pattern: Modal + checkboxes + action buttons mirrors UI patterns physicians are already used to

Final flow: View and edit a note (prototype)

Focus Area 3

Focus Area 3

Focus Area 3

Bonus design: Future dashboard for physicians to stay on top of notes and daily tasks

Bonus design: Future dashboard for physicians to stay on top of notes and daily tasks

In the user interviews, we also saw physicians' need for an easier way to review, locate, and follow up on patient notes alongside their daily schedule.

While the inbox redesign and note editing flow tackled immediate pain points, I also had some time left in my internship to think ahead. After discussing with the founding team, I designed a dashboard concept that could give physicians a high-level view of incomplete tasks and quick access to patient notes, helping them stay organized and proactive throughout the day.

Dashboard concept

Takeaways.

Takeaways.

Takeaways.

Work within constraints. Design for the capability

Work within constraints. Design for the capability

During my internship, the greatest challenge was the limited time frame for organizing complex information and creating experience for a new feature. The team expected me to quickly integrate into the design process. It forced me to prioritize and make efficient decisions to deliver the MVP effectively.

While it was tempting to explore all possibilities of new features that could enhance the overall user experience, I had to ensure that the designs I proposed were feasible within the team's resources and capabilities. This experience taught me the importance of aligning design ideas with project constraints and making informed decisions that balance ideation with practicality.

Device-specific characteristics and user behaviors

Device-specific characteristics and user behaviors

Recognizing the importance of tailoring designs to specific devices, I learned to adapt my approach to accommodate user preferences and interactions on each platform. This insight has significantly influenced my subsequent design projects, as I now consider device-specific considerations to create optimal user experiences.

Recognizing the importance of tailoring designs to specific devices, I learned to adapt my approach to accommodate user preferences and interactions on each platform. This insight has significantly influenced my subsequent design projects, as I now consider device-specific considerations to create optimal user experiences.

© 2026  Xiaochen(Iris) Shi

© 2026  Xiaochen(Iris) Shi

© 2026  Xiaochen(Iris) Shi

© 2026  Xiaochen(Iris) Shi