🎙️ Note Assist - AI Medical Scribe

An EHR Native, AI-powered audio capture app that transcribes and summarizes clinical notes.

Overview

Note Assist is Elation Health’s first AI product and flagship generative AI feature, an in-chart medical scribe that captures clinical encounters in real time. Originally “AI Scribe,” a 2023 internal hackathon winner, it became a full-scale product in 2024. Note Assist officially launched in the US in 2024. We launched first to primary care, then expanded to other specialties through prompt tuning and settings.

My Role

I co-created the original AI Scribe concept during a 2023 internal hackathon, designing the prototype that became Note Assist. I led early research, information architecture, and interaction design for the first user rollout.

The Challenge

Clinicians were burning out from documentation. Many spent nights and weekends finishing visit notes, sometimes forgetting critical follow-ups like referrals or lab orders.
AI scribes existed, but most:

• Produced generic, incomplete notes that required heavy editing

• Lacked EHR integration, forcing manual copy-paste

• Had no access to patient context, reducing clinical accuracy

Context: Why Now

• 62% cite “excessive documentation” as their top burnout driver… [read]

• EHR after-hours work makes clinicians 2.4× more likely to burn out… [read]

• AI scribes cut cognitive load, save time and improve patient presence…[read]

This was the perfect moment to deliver a fully embedded, AI-powered documentation tool that fit directly into Elation’s native workflows.

Design & Product Goals

• Win clinicians’ trust with real-time transcripts and verifiable AI output

• Eliminate busy work by capturing directly inside the EHR

• Deliver instantly so notes are ready before the provider is

• Make it clinically sharp with patient-aware specialty-tuned prompts

• Cut pajama time charting and missed follow-ups

• Keep UI light, movable, and language-aware

• Lock down privacy from first word to saved note

Iteration #1. Mapping the Experience

Before touching pixels, I mapped the end-to-end flow, every screen, every decision point, and the kind of experience we wanted clinicians to have.

My first pass kept things familiar: show the AI summary in the Scribe dialog, let users edit, then export to the visit note. But when we asked, “What actually makes this different from every other scribe on the market?”, the answer was obvious, Elation’s native integration.

We stripped away the extra handoffs. Instead of parking the summary in a separate dialog, we pushed it directly into the visit note in real time. Start recording, pause/resume if needed, end the session, the note updates automatically. No export step, no copy-paste, no context switching.

Later iterations went further. We improved latency so the visit note updated continuously throughout the encounter. Clinicians could see live, in-context documentation as they spoke without breaking their flow. I also designed the dialog to be scalable, with space for future settings and drag-and-drop repositioning to avoid blocking critical EHR data.

Iteration #2: Testing, Feedback and Refinement

The first prototype went to internal stakeholders and “co-design” clinicians, customers willing to help shape the product. Their feedback led to three big improvements:

• Bigger transcript area: Easier mid-visit scanning, and for some providers, a real-time aid when speaking with patients with accents or speech impairments.

• Real-time audio feedback: An equalizer showing mic pickup levels, so providers knew their voice was being captured clearly.

• Systematic error handling: Categorized every possible failure mode and built consistent, reusable resolution patterns instead of one-off fixes.

We are creating a scribing experience that is fast, integrated and most importantly trustworthy.

Iteration #3: Speed, Intelligence and What’s Next

By the third iteration, speed was no longer a nice-to-have, it was core. The visit note now updated continuously during the encounter, with near-instant latency.

The team also added Actionables: AI-generated to-do items pulled straight from the transcript: place an order, write a referral, prescribe a medication, all collected in their own tab.

Providers could check them off once completed, closing the loop and reducing missed follow-ups.

Other major upgrades included:

• Multi-language support: First with Spanish and Chinese, expanding to serve more patient populations.

• Granular customization: Ability to exclude certain Visit Note sections from scribing, work with existing templates, and even rewrite only selected sections based on both the transcript and existing chart data.

• Tighter EHR integration: Planning for the next-gen Visit Note where Scribe is embedded directly into the charting surface, rather than living in a floating dialog.

Results & Impact

Note Assist’s launch proved that deeply integrated, clinician-first AI can transform documentation.

Key outcomes:

• 13 minutes saved per visit on average, adding up to hours reclaimed every week for patient care instead of paperwork (elationhealth.com).

• 75% increase in reported “joy” and 87% reduction in burnout among clinicians using Note Assist (vbdata.cn).

• Actionables feature helped 65% of clinicians reclaim 5–30 minutes per patient, while 70% reported feeling more focused and engaged during visits (elationhealth.com).

• Now available in 13 languages, supporting diverse patient populations and improving accessibility of care (elationhealth.com).

Final Reflection

• As Elation’s AI portfolio grew, our design team created a unified AI design language, consistent visuals, interactions and terminology across products. Note Assist adopted it in later iterations, making the experience clearer and more intuitive for clinicians.

• We added a lightweight feedback widget right in the workflow, letting clinicians quickly rate each session and leave comments. This kept our feedback loops sharp, so we could follow up immediately after both glowing and critical sessions.

• AI expectations rise fast. What impressed at launch became the baseline within months. That pushed us to move quickly, fine-tuning prompts, testing new models like ChatGPT, Claude, and Anthropic, and steadily improving quality.

• Designing Note Assist reminded me that great AI isn’t just about polished UI. It’s about trust, seamless flow, and measurable clinical impact. It became a true partner in the patient encounter. Building something that’s both technically advanced and human-centered is exactly the kind of work I want to keep scaling.