Voice-to-Chart AI Assistant for Home Healthcare

My role

Product designer (iOS/iPadOS)

SCOPE

Zero-to-one B2B SaaS initiative

Company

nVoq is a cloud-based speech recognition platform that helps physicians and clinicians streamline documentation and automate workflows using voice. nVoq empowers healthcare organizations to reduce administrative burden, cut down documentation time, and shift provider focus back to direct patient care.

The PROBLEM

Home health clinicians (nurses, physical therapists, and occupational therapists) spend a massive chunk of their day battling documentation rather than caring for patients.

The Documentation Burden: After a visit, especially complex Start of Care appointments requiring exhaustive OASIS assessments with hundreds of questions, clinicians face hours of manual EHR data entry.
Compromised Human Connection: To capture every necessary detail in real time, clinicians are often forced to juggle screens and typing during home visits, taking their attention away from the patient and eroding trust.

The GOAL

Build a HIPAA-compliant mobile companion app that allows clinicians to record visit conversations freely, shifting their focus back to human connection. The app handles the heavy lifting through asynchronous transcription, summarization, and AI-powered form pre-filling, while maintaining strict clinical oversight.

Accelerating Design with AI Tools

To move rapidly from concept to high-fidelity validation, I utilized an AI-augmented design workflow. By leveraging AI tools like Claude and Figma Make during the exploration and prototyping phases, I rapidly iterated, tested, and refined interaction flows ahead of user sessions with field clinicians.

CORE USER JOURNEY

From ambient recording to a verified OASIS submission

To make the app's value concrete before diving into the design philosophy, this section walks through the primary user flow. By moving from a live home visit to a finalized, clinician-verified EMR submission, the experience is built to eliminate hours of post-visit desk work while keeping the user fully in control.

1. HOME PAGE & RECORDING INITIATION

The Visit-Based Home Screen: Clinicians land on a home screen that serves as their central dispatch, displaying all upcoming visits. Each appointment is presented as a clear card detailing critical patient info, such as the patient's name, Date of Birth (DOB), Medical Record Number (MRN), visit type, and a prominent "Record" action button.
Consent & Context Check: Tapping record triggers a vital pre-recording screen that prompts the clinician to confirm patient identity and verify they have received permission from all participants to record the audio conversation.
The 3-Second Buffer: A 3-second countdown gives clinicians a moment to collect their thoughts, put down their phone, and settle into the conversation before capture begins, with an easy out via a cancel button if needed.

2. LIVE RECORDING & AMBIENT BACKGROUND CAPTURE

Active audio verification: During recording, a responsive live waveform and timer provide immediate visual feedback that audio is being captured properly. Clinicians have full control with pause, cancel, and end-recording actions.
Lock screen: To ensure clinicians are never trapped staring at an active phone screen during a visit and to preserve battery, the recording architecture fully supports locked-screen background capture. Furthermore, a dedicated Live Activity widget on the lock screen and notification center gives clinicians continuous, at-a-glance visibility into the active timer and waveform without needing to unlock their device.

3. AI SUMMARY & TRANSCRIPT

Summary and transcript: Once recording ends, the audio is processed to generate a comprehensive AI summary alongside a full, time-stamped transcript that intelligently detects and labels individual speakers. Clinicians can also edit speaker labels if necessary, ensuring full control over attribution, and have direct access to the raw audio recording for playback and verification whenever needed.
Advanced note refinement: To streamline summary editing and formatting, clinicians can leverage powerful built-in nVoq tools:
- Note Assist: An intelligent checklist tool that ensures all necessary clinical talking points and compliant categories are included for the selected note type.
- Note Compose: Automatically formats raw text into structured clinical formats (such as SOAP notes).
- Shortcuts: Allows clinicians to instantly insert custom note snippets for rapid documentation.
  .

4. OASIS FORM FILL & SUBMISSION FLOW

Section Sorting: Once the summary is reviewed, tapping "Run review" pre-populates the exhaustive assessment form such as OASIS, using the inferred information from the summary and transcript. To optimize clinician review time, questions are intelligently sorted with sections that need attention first.
Submission guardrails: The interface uses confidence indicators and color-coded pills (green, amber, red) to flag items. Critical missing or low-confidence items completely block submission, while the rest of the items require manual review or confirmation before the final submit button unlocks. This guarantees that no data is blindly submitted without explicit clinician sign-off.

DESIGN STRATEGY & ETHICAL AI

Balancing automation with accountability

In home healthcare, particularly during high-stakes Start of Care visits governed by exhaustive OASIS-E assessments, accuracy is non-negotiable. While the goal of the AI assistant is to eliminate manual documentation fatigue, my design philosophy was rooted in a core principle: the AI does the heavy lifting, but the clinician maintains absolute ownership and accountability. The system is designed strictly to help clinicians remember details and streamline documentation, ensuring that clinical judgment always remains in human hands. Throughout every stage - from recording to form pre-filling - the app empowers clinicians to review and modify data while remaining fully aware that AI can make mistakes.

TRANSPARENT AI REASONING

To build trust, features like the AI Reasoning drawer give clinicians a clear window into why the system made a specific suggestion, complete with direct transcript quotes and source tags. In cases where the AI detects ambiguity or identifies multiple potential answers, it surfaces the competing options side-by-side (complete with evidence sources and confidence scores) rather than making a blind guess, leaving the final choice to the clinician.

DELIBERATE FRICTION

Feedback from EHR quality specialists during user testing heavily reinforced that users should not be allowed to blindly breeze through pre-filled data. To protect against errors, future disputes, and liability, thoughtful friction and confirmation safeguards are integrated at key milestones throughout the workflow:

Consent verification: Requiring explicit confirmation of patient identity and recording consent before capture begins.
Action confirmations: Prompting warning or confirmation pop-ups for irreversible or final actions, such as ending a recording.
Mandatory review: Enforcing review workflows prior to submitting the final form into the EMR, serving as a digital verification signature that reinforces clinician responsibility.

DESIGN FOR EDGE CASES

Designing for unpredictable environments

Home health clinicians operate in dynamic, real-world environments where reliable internet connectivity is never guaranteed and patient privacy is paramount. Designing for the field meant building an architecture capable of handling abrupt network drops while safeguarding sensitive medical information.

OFFLINE RECORDING & PROCESSING

Because cell coverage can drop unpredictably while clinicians travel between patient homes, the app allows full audio recording even when completely offline. Once internet connectivity is re-established, queued recordings automatically feed into the AI transcription and processing pipeline without requiring manual intervention. The user is alerted via in-app and push notifications when their visits are done processing

HIPAA COMPLIANCE & DATA SECURITY

To maintain strict HIPAA compliance and data security, patient data is stored securely in the cloud rather than retained permanently on the device. Data is automatically removed from local storage after a certain period of time, but can easily be redownloaded from the cloud when the device is back online. Additionally, to protect sensitive patient records from unauthorized access, the app requires users to periodically re-verify their credentials through a passcode, fingerprint, or face unlock after periods of inactivity. Furthermore, to prevent accidental privacy violations during playback, the system detects when an external Bluetooth speaker connects and displays an explicit warning alert, requiring the clinician to manually confirm their environment is safe before any audio can be played.

Validation & user testing insights

Testing experience with field clinicians and EHR compliance experts

Before finalizing the product experience, the prototype was put through rigorous user testing sessions with real home health clinicians (including registered nurses and physical therapists) as well as EHR quality and regulatory compliance specialists. These sessions provided invaluable qualitative validation for a zero-to-one product footprint. The following key insights were found:

OWNERSHIP OVER AUTOMATION

Testing confirmed that clinicians strongly welcomed technology to streamline their heavy documentation burden, but they universally insisted on maintaining final control. They viewed the AI as a helpful cognitive assistant rather than an autonomous decision-maker, emphasizing the absolute necessity of being able to review, verify, and edit outputs.

INTENTIONAL FRICTION = GOOD

Feedback from an EHR quality specialist validated the inclusion of deliberate friction points and mandatory review gates. Rather than treating validation steps as friction, compliance experts welcomed safeguards that prevent users from blindly bypassing reviews, framing it as an essential protection against documentation errors and liability.

INTUITIVE NAVIGATION

Despite handling complex background states, audio pipelines, and detailed clinical assessment frameworks, participants found the core mobile UX to be highly intuitive and natural to pick up during fast-paced clinical workflows.

Next steps

What's next?

As a zero-to-one product footprint established through iterative testing, the current application lays a robust foundation for automated ambient documentation. However, looking ahead, several powerful features and architectural extensions are planned to further streamline clinical workflows and expand product capability.

REAL-TIME VISIT GUIDE

Leveraging the visit type pulled directly from the EHR (such as a Start of Care appointment), a dynamic real-time checklist will be introduced to analyze conversations as they happen, flagging potentially missed assessment items before the clinician leaves the patient's home.

MULTI-RECORDING MANAGEMENT

Expanding beyond the current single-recording limitation per visit to fully support multi-recording workflows, allowing clinicians to seamlessly capture, separate, and manage pre-visit notes, main visit interactions, and post-visit addendums.

HISTORICAL PATIENT DIRECTORY

Implementing a dedicated patient directory view enabling clinicians to review past visits for individual patients at a glance, complete with high-level progress summaries over time.

Let's chat!

I'd love to connect with you! Let's talk about anything, and grab some coffee

Send me a message