ClaimClear: Medical Billing Tool

Skills Applied

Design | Research | User Testing | AI-Assisted Design

Softwares Used

Figma | Claude | Maze | FigJam | Netlify

User/Consumer Base

B2B | B2C

Industry Experience

Healthcare | HealthTech | MedTech

The Problem

Healthcare billing fails everyone before anyone even walks out the door.

Patients arrive without knowing what care will cost. When the bill comes, 90% can’t interpret the codes on it, and 100% react emotionally to surprise charges. So they delay or skip care. They call the billing department in a panic.

On the other side, billing admins submit claims without pre-visit error checks and 1 in 8 get denied. Each denial costs $57.23 in rework, adding up to $48 billion in losses across U.S. hospitals in 2025 alone.

The rules exist to fix this. HIPAA’s Right of Access, the No Surprises Act, and CMS price transparency rules all require providers to give patients clearer cost information upfront. In practice, enforcement is spotty and the gap remains wide open.

ClaimClear was built to close the gaps before a claim is ever submitted.

What is ClaimClear?

ClaimClear is a pre-submission billing tool for healthcare providers.

Before a claim is ever filed, it gives billing administrators a streamlined workflow to review AI-suggested CPT codes, catch bundling and modifier errors, and send patients a plain-language cost estimate all in one place.

The result: fewer denials, less rework, and patients who arrive informed instead of blindsided.

Who it’s for

Billing Admins: Catch errors before submission, cut rework time, reduce denials.

Patients: Understand costs upfront and make informed decisions about care.

Providers: Meet No Surprises Act and CMS price transparency obligations without extra admin lift, and potentially see fewer appointment cancellations from patients caught off guard by cost.

Discovery & Validation

The project started with social listening, not a survey. I scoured Reddit threads where patients vented about surprise bills and billing admins complained about denial rework. I was looking for recurring, specific pain points rather than assumptions.

To validate what I found, I ran a 10-person survey on Maze. The results confirmed the pattern:

90%

100%

50%

One response stuck with me:

“I felt slightly overwhelmed and annoyed. I logged into my medical portal and tried to figure out why it was higher than expected.”

The pattern was clear on both sides: patients had no way to understand costs before they hit, and admins had no way to catch errors before claims went out. ClaimClear’s core workflow, pre-visit cost estimates and pre-submission error flags, came directly from this gap.

Designing with AI as a Collaborator

This is where the project moved fastest and where I leaned hardest on AI as a working partner, not a shortcut.

Ideation. I used Claude to think through information architecture and flow options based on the discovery findings.
First mockup. I connected Claude to Figma via MCP and had it draft an initial low-fidelity design directly in Figma.
Design system. I refined that draft manually and built out a proper design system with typography and color tokens as well as a component library.
Re-skin. I directed Claude to apply the design system to the existing mockup, then refined the result manually until it matched.
Code. Once the core flow felt right, I had Claude build it out as a working HTML/CSS/JS prototype.
Ship it. I asked Claude how to make the prototype public and it walked me through deploying on Netlify, which I used to get a live, testable link.
Iterate. From there, I went back to Figma to direct further design changes, which I’d hand to Claude to implement in the live HTML file, something I repeated until the flow was ready for user testing.

Throughout, AI accelerated execution, but every design decision and refinement was mine. Claude built what I directed; I never shipped a first draft.

Built with Compliance in Mind

Healthcare data and billing are heavily regulated, so before designing the workflow, I completed a HIPAA for Business Associates certification (IACET-accredited, 360training/HIPAA Exams) to understand the legal landscape ClaimClear operates in.

That research shaped specific product decisions:

Regulation	What it requires	How ClaimClear addresses it
No Surprises Act	Good Faith Estimates before care	Generates patient-facing cost estimates pre-visit
CMS Price Transparency	Clear, accessible pricing (enforced with fines up to $309,738 in 2025)	Surfaces CPT codes and costs in plain language
HIPAA Right of Access	Timely patient access to billing/records info	Designed to give patients visibility before, not after, the bill arrives

Enforcement of these rules is still inconsistent with over 1,000 information-blocking complaints filed since 2021 with no penalties issued yet. That gap is part of the opportunity: providers who get ahead of it reduce both compliance risk and patient frustration.

Next Steps

Lessons Learned

Designing ClaimClear taught me that AI works best as a collaborator, not a shortcut. The strongest results came from directing it with clear intent, then refining by hand. Pairing social listening with quantitative validation also proved essential: Reddit surfaced the pain points, the Maze survey confirmed they were real and widespread.

Continued Iteration

I’m continuing to test ClaimClear with real patients and billing admins, sharpening the flow based on what breaks. I’m also using this project to keep deepening my AI collaboration skills, pushing further into AI-assisted design and development workflows.

Y Combinator

To bring ClaimClear beyond a prototype, I’ve set up a profile on Y Combinator’s cofounder-matching platform (currently pending approval), looking for a technical cofounder to help build it into a real product.