Real results. Across industries.

Aging patient balances. Front-desk staff chasing manually. Inconsistent follow-up. Legacy billing platform plateauing at ~53% of patient responsibility collected at 90 days — roughly the industry average per HFMA MAP Keys. The fastest-growing segment of healthcare AR, still being chased by hand.

AI-assisted patient-billing operating system with 12 observable capabilities: approval-gated settlement campaigns, inbound SMS intent classification, autonomous safe-intent replies, dispute manager, risk scoring + worklist prioritization, payment prediction, send-time optimization, payment-plan negotiator, and a semi-autonomous follow-up agent. Built on Retell AI + Twilio + Stripe + Make.com, with a full operational export pack for reproducible measurement.

60 days from concept to production. 123 days of post-launch measurement published.

December 15, 2025

$535,623 collected in 123 days — +129.4% vs. the legacy platform’s equal-window baseline. $182.62 per paying account (industry top-quartile per HFMA). 75.8% of patient responsibility collected by 90 days vs. 53.1% on legacy (and ~50–60% industry median). Cost to collect: 1.39% — below HFMA’s 2–3% best-practice benchmark. Full methodology, sources, and honest caveats in the deep-dive case study.

Most patient balances never get chased well because the economics of manual phone-and-mail don’t work. The platform collapses the operating cost to ~1.4% of collections while hitting top-quartile industry KPIs — on a live patient panel, measured over 123 days, with every number reproducible from raw exports.

$3M agency running entirely on spreadsheets. Owner spent ~80% of his time managing data instead of clients. Duplicate entry across 3 systems. No central source of truth. Jay couldn’t tell me what he needed in software terms — but he didn’t have to. He walked me through his operation, and I pattern-matched the fix on the spot.

An in-person working session, not a product. Jay walked me through his workflow — spreadsheets, duplicate entry across three systems, client data in his head, nothing centralized. I put the pieces together in real time and identified what the business actually needed: not a CRM rollout, not a Salesforce implementation, not off-the-shelf. A fully custom CRM built from scratch for his exact workflow. Scope was set in the conversation.

A fully custom CRM, built from scratch for Avva’s specific workflow — no off-the-shelf platform, no SaaS template. Architecture scoped in-person at the facility, data migrated from spreadsheets, workflow automation wired, code walked through with the owner so he could maintain it, then full GCP deployment end-to-end. Team trained before we left.

~24 hours of build time — from first line of code to live on GCP. Not wall-clock; compressed work cycles across a working day and a half, with scope self-evident from the scoping session.

The operator’s lens catches in a conversation what a consultant bills weeks of “discovery” for. Jay didn’t walk in with a scope or a brief — he walked me through his operation, and I pattern-matched the fix in real time. Most Build engagements run 4–12 weeks because the first chunk is discovery. We compressed discovery into the meeting itself because I’d lived the same pattern before. That’s the difference between hiring a consultant who needs weeks to understand your business and hiring an operator who’s been the one trapped inside a business running on spreadsheets.

Paper intake, PDF forms, manual data entry, HIPAA exposure.

End-to-end digital intake platform on Firebase + Twilio + AWS SES with Gemini AI summarization.

Active in production at imaging center.

Patients arrive with forms pre-completed. Repeat visitors skip redundant fields. Staff data entry eliminated. HIPAA-compliant end to end. Published performance metrics coming Q2 2026.

Intake is the first 15 minutes of every patient experience. Automating it frees staff for work that requires judgment.