Loading...

From Voice App to Enterprise AI Platform

How I Led UX Design for an AI-Powered EHR Platform That's Helping to Transform Priorities Across a $100B Private Equity Portfolio

← Back to Home 

The Challenge: Disrupting Decades-Old EHR Monopolies

Here's what most people don't realize about healthcare technology: the major EHR platforms are monolithic code nightmares built decades ago by companies that add features, not create products. When Greenway Health - part of Vista Equity Partners' $100B portfolio - decided to build something different, they weren't just creating another voice transcription tool. They were architecting the future of healthcare.

The initial vision seemed straightforward: help burned-out physicians reclaim their "pajama time" through ambient voice technology. But as Lead UX Designer, I saw the bigger opportunity. This could evolve from a simple sidekick app into a full AI-powered EHR platform - one that could serve Vista's 500 million users across 90+ enterprise companies.

The stakes were massive. Healthcare providers were drowning in documentation, spending 6 hours daily on EHRs with 2.5 hours bleeding into personal time. Chronic underbilling was costing practices millions. Providers were missing federal MIPS money due to poor eCQM documentation. And 80% of our patient portal traffic was mobile - yet most EHR platforms barely worked on phones.

Note: Screenshots shown are design mockups as the actual platform remains proprietary and in beta.

Medical App Smartphone Streamline Icon: https://streamlinehq.com

Industry

Healthcare Technology | AI-Powered EHR Platform | Enterprise Software
Voice Id User Streamline Icon: https://streamlinehq.comvoice-id-user

My Role

Lead UX Designer orchestrating 100+ stakeholders across 6 development teams
Professions Woman Doctor 2 Streamline Icon: https://streamlinehq.comprofessions-woman-doctor-2

Impact

Platform secured 8-figure funding, saves providers 500 hours annually, designed for 2+ dozen languages
Love Heart Hold Streamline Icon: https://streamlinehq.comlove-heart-hold

Scale

Platform processing billions of healthcare records, serving hundreds of providers with 638 practices migrated

Discovery: Understanding Every Healthcare Persona

Building medical platforms means designing for seemingly countless personas - from front desk staff juggling appointments to billing specialists fighting denied claims to providers making life-or-death decisions. Each persona needed what felt like a completely different product, yet everything had to work seamlessly together. An additional challenge was the fact that while our development teams were building the first iteration that was fundamentally pretty simple, the design team had to work months or years in advance.

Our research approach combined traditional UX methods with cutting-edge analytics:

  • Clinical Shadowing: 45+ hours across specialties understanding real workflows
  • Stakeholder Interviews: 30+ providers, plus regulatory, compliance, and billing teams
  • Data Analysis: Pendo session replays revealing rage clicks and dead zones
  • Journey Mapping: 100+ daily schedule patterns across diverse user types
  • Focus Groups: Regular sessions with blunt provider feedback (they don't hold back)

Using Pendo and Domo, I tracked every click, every hesitation, every abandoned workflow. One discovery helped me to build a case to change an issue that I had concerns with previously. An icon we'd hidden behind a menu was being accessed more than some of our surface-level features. Real data beats assumptions every time.

Journey maps highlighting pain points
Yeah yeah I know, journey maps in a portfolio case study... how original, right? But this step wasn't about building pretty visuals. It was about truly understanding the user's pain points across all aspects of the platform and the various workflows.

From MVP to Platform: The Evolution Story

What started as a voice transcription app has grown 4-5x in scope. Here's how Clinical Assist evolved:

Phase 1: The Voice Foundation

The initial app was elegantly simple - record patient conversations, generate structured SOAP notes. Working with Nabla's AI, we designed interfaces that processed natural conversation through 2+ dozen languages, handled medical jargon, and structured unorganized data into billable documentation. In just the first few months, providers sent over 25,000 structured visit notes from the app to their EHRs.

Phase 2: The Platform Vision

Success bred ambition. Leadership saw the potential for something bigger - a full EHR platform built AI-first from the ground up. Now I'm designing complete patient charts, appointment management, revenue cycle optimization, clinical decision support interfaces, multi-language workflows, predictive analytics dashboards, and federal compliance tools.

Phase 3: The AI Integration

We're not just adding AI features - we're rebuilding healthcare workflows around AI capabilities: intelligent summarization UX, voice-first navigation, automated billing optimization, and smart notification systems. Every workflow is designed to leverage AI's strengths while maintaining human oversight. That meant building prompt playbooks for each persona and pairing them with the infrastructure guardrails you'll see next.

Mobile-First in Healthcare: A Paradigm Shift

I worked closely with development teams to define UX workflows that determine local storage needs and error handling. When you're designing for a provider in a rural clinic with spotty internet, every offline scenario matters. We mapped out:

  • Sync state indicators showing what's saved locally vs. cloud
  • Conflict resolution workflows for when connections restore
  • Progressive data loading prioritizing critical patient information
  • Graceful degradation ensuring core functions work offline

We've seen a massive shift toward smaller devices in healthcare. With 80% of patient portal users on mobile, and providers increasingly using tablets and phones for bedside care, mobile-first isn't optional - it's survival.

Designing the AI Brain: Prompts, Guardrails & Infrastructure

Scaling from a humble proof of concept to an enterprise platform meant treating AI orchestration like part of the design system. I spent as much time shaping prompts and data pipelines as I did polishing screens so every insight surfaced in the UI felt trustworthy. While it's gaining mainstream traction, there is still a palpable trust issue in healthcare when talking AI.

Prompt Systems Built for Clinicians

Clinicians don't talk like chatbots. They speak in shorthand, acronyms, and context that can shift from visit to visit. I built prompt frameworks that respected that reality:

  • Role-based prompt wrappers tuned for triage nurses, primary care providers, and specialists.
  • Context injection pulling in chart data, regulatory requirements, and visit intent so AI outputs stayed compliant.
  • Real-time feedback loops capturing provider edits to reinforce what "good" looks like across personas.

Shipping Without a Manual

AWS was shipping new capabilities faster than they could document them. Bedrock's evolving documentation still pointed back to older SageMaker examples, so we treated ourselves as the documentation team.

  • Reverse-engineered behaviours: logging every odd response and publishing internal playbooks so engineering, product, and QA stayed aligned.
  • Weekly working sessions with AWS solution architects to validate guardrails, security models, and fallback plans.
  • Design/dev toggles that exposed infrastructure readiness inside the UI, keeping feature flags and clinician expectations in sync.

With prompts, guardrails, and AWS infrastructure hardened, our demos felt less like vaporware and more like software already operating in clinics. That credibility set the stage for the funding story that follows.

The 8-Figure Funding Story

Here's something most case studies won't tell you: sometimes your designs directly influence massive funding decisions. When Vista Equity Partners was evaluating investment in our AI vision, I created:

Video Player Monitor Streamline Icon: https://streamlinehq.com

AI-Assest Vision Videos

4 minutes of footage showing our 3-year platform roadmap
Mobile Qr Code Streamline Icon: https://streamlinehq.com

Interactive Prototypes

Demonstrating automated workflows from diagnosis to billing
People Woman 5 Streamline Icon: https://streamlinehq.com

Character Alignment

Using Flux, Veo, and Kling to maintain consistency across clips
Presentation Statistics Streamline Icon: https://streamlinehq.com presentation-statistics

Stakeholder Presentations

Directly shown to Vista's founder during funding negotiations

These materials helped secure 8-figure funding for the platform. Not bad for a sprint's worth of work with a few AI image & video tools and After Effects. The videos demonstrated how our platform could automate everything - scheduling, billing, personalized patient communications - all the pain points destroying healthcare today.

This wasn't just about pretty visuals. It was about translating a complex technical vision into something a PE firm founder could immediately understand and believe in. The prototypes I'd been refining for 2-3 years finally had their moment, showing how healthcare could work in a highly automated, AI-driven future.

The Results: Transforming Healthcare at Scale

The metrics tell the story, pulled from third-party validation and our own analytics:

Voice screens, Appointments, Note Draft, and Transcribing real-time
From today's appointments and patient flowtracking, to intelligent, natural language processing that not only captures and trascribes the conversation (even in different languages), but then builds a structured clinical note automatically. Something not right? Or do you prefer bullet points? The LLM can edit it even further. (Note: These are early unused mockups)
Time Reclaimed Per Year
500hrs
Avg Time Saved Per Day
2hrs
Patients Report More Attention
81%
Clinicians Feel Less Pressure
90%

Downstream impacts even more impressive:

By improving documentation quality, we're helping providers:

  • Identify missed diagnoses through AI-powered chart review interfaces
  • Improve patient satisfaction through more present, engaged care
  • Reduce denied claims through better coding accuracy
  • Capture MIPS and HCC dollars previously lost to poor tracking
  • Reduce support tickets by 65% through intuitive, self-service design

Additional Metrics:

  • Documentation Time Saved Daily: 2+ hours
  • Languages Supported in Design: 2+ dozen and growing
  • Practices Migrated: 300+ and growing
  • Structured Visit Notes Sent: 25,000+ in first 90 days
  • Healthcare Records Processed: 9.5 billion FHIR resources

“Before using this tool, I was juggling documentation, listening, and recalling patient details. Now, I’m more engaged and focused during appointments, as the tool handles documentation seamlessly. It’s freed me up to connect more meaningfully with my patients.”

Why This Matters: The Bigger Picture

This isn't just about making software prettier or workflows faster. We're fundamentally changing how healthcare works. When providers save 2 hours daily on documentation, that's 2 hours returned to patient care. When our interfaces help providers see AI-caught diabetes indicators, that's a life potentially saved. When our translation workflows help a provider understand a patient's symptoms in their native language, that's health equity in action.

The platform is now Greenway's flagship product and highest priority. It's leading the entire organization's transformation into an AI-centric company. Other portfolio companies under Vista are watching and learning. We're not just building features - we're setting the standard for what healthcare technology should be. Our own CEO Pratap Sarker publicly declared that “all new innovations from Greenway will be underpinned by generative AI moving forward.”

The Design Process: Managing Complexity at Scale

Monitoring & managing design across a half dozen development teams and 100+ seats & stakeholders requires more than just Figma skills.

Stakeholder Alignment

  • C-Suite to Clinic: Jumping from 10,000-foot views with directors to pixel-level CSS discussions with developers
  • Clinical Translation: Working with our CMO and clinical staff to ensure medical accuracy
  • Compliance Navigation: Every design decision filtered through regulatory, legal, and HIPAA requirements
  • User Story Management: Maintaining single sources of truth across dozens of parallel workstreams

Data-Driven Iteration

  • Pendo Analytics: Tracking every interaction, identifying friction points through rage clicks and dead zones
  • Domo Dashboards: Visualizing usage patterns to inform design decisions
  • Focus Group Insights: Regular sessions with providers who give brutally honest feedback
  • A/B Testing: Data-driven decisions on everything from icon placement to workflow order

AI-First Design Principles

  • Confidence Indicators: Visual cues showing when AI suggestions need human review
  • Transparent Processing: Users always know when AI is working and what it's doing
  • Fallback Workflows: Graceful degradation when AI features aren't available
  • Learning Loops: Interfaces that show how the system improves with provider corrections

Looking Forward: The Future We're Building

Clinical Assist continues to evolve. Currently in development:

  • Predictive Clinical Intelligence Dashboards: Interfaces for AI-identified care gaps and chronic condition risks
  • Automated Prior Authorization Workflows: Reducing administrative burden through intelligent form design
  • Patient-Initiated Scheduling UX: AI-powered symptom matching and appointment routing
  • Continuous Learning Indicators: Showing providers how the platform adapts to their practice patterns

But perhaps the most important evolution is cultural. We're showing an entire industry - one notorious for resistance to change - that AI can enhance rather than replace human connection. When a provider tells me "I can finally look my patients in the eye again," that's when I know we're succeeding.

The Lesson: Design at Scale Requires More Than Pixels

This project taught me that Principal-level design isn't just about creating beautiful interfaces. It's about:

  • Orchestrating complexity: Managing teams and stakeholders across multiple teams
  • Thinking in systems: Designing for millions of interactions across diverse platforms
  • Proving value: Using prototypes and vision videos to secure 8-figure investments
  • Bridging worlds: Translating between clinical needs, technical constraints, and business goals
  • Leading through influence: Driving change across an organization without formal authority

When you're designing AI's future for not only your ORG, but a $100B private equity portfolio, every decision impacts millions of lives and dollars. Every workflow affects hundreds of providers. Every optimization can mean the difference between burnout and sustainability for our healthcare system. It's a lot of pressure. But a task I hold dear to my heart as I have a number of healthcare workers that I know and love - including my wife. And knowing that I can help to improve the tools they use everyday to save lives is not something I take lightly.

That's the kind of challenge that gets me up in the morning.

Want to learn more about this project? Read Greenway's announcement about Clinical Assist

Home | Next Project →