Amby Health Is Replacing Your Ambulance Agency's Back Office With AI

The Macro: EMS Billing Is a Mess Nobody Talks About

Here’s something most people don’t think about: when an ambulance picks you up, the billing process that follows is one of the most convoluted workflows in all of healthcare. And healthcare billing is already the gold standard for convoluted workflows.

Every ambulance run generates a Patient Care Report. That report needs to be reviewed for medical accuracy, coded for billing, checked for compliance with Medicare and Medicaid regulations, assessed for liability risks, and submitted to payers with the right documentation. Miss a modifier code? Claim denied. Document the wrong level of service? That’s a potential fraud flag. Fail to note medical necessity for transport? Good luck getting paid at all.

Most ambulance agencies handle this with large back-office teams. Billing specialists, quality assurance reviewers, compliance officers. A mid-size agency running 50 to 100 calls per day might have 10 to 15 people whose entire job is processing the paperwork generated by those calls. The median time from ambulance run to clean claim submission is measured in days, sometimes weeks. Revenue sits in limbo while humans manually review charts and argue about ICD-10 codes.

There are existing players in this space. Digitech and ImageTrend handle electronic patient care reporting. Zoll provides billing and data management. Intermedix (now Guidehouse) offers revenue cycle management for EMS agencies. But most of these tools digitize the existing manual workflow rather than fundamentally changing it. They replaced paper forms with digital forms. The human review bottleneck remains.

The Micro: MIT Dropouts Who Know Both AI and Healthcare

Amby Health is building an AI system that analyzes patient care reports in real time, automatically codes claims, identifies documentation errors, flags liability risks, and ensures compliance. The pitch is simple: replace the back-office team with AI that does the same work in seconds instead of days.

The founding team brings a specific combination of skills that matters here. Yos Wagenmans is CEO. He’s an MIT dropout who worked as an engineer at Meta, did ML research at MIT CSAIL, and built pre-hospital software at ALLM, a Japanese emergency medicine tech company. That last one is key. He’s actually built software for EMS before. Timmy Dang is CTO, also an MIT dropout, with engineering stints at Amazon and Bloomberg, AI research at MIT Media Lab, and healthcare data science work at OM1. They’re a two-person team in San Francisco, part of YC’s Winter 2025 batch with Harj Taggar as their primary partner.

What sets this apart from generic “AI for healthcare billing” is the domain specificity. EMS billing is not hospital billing. The codes are different. The compliance requirements are different. The documentation standards are different. Medicare reimbursement rules for ambulance services are their own specialized nightmare. Building AI that understands the difference between a BLS emergency transport and an ALS2 intervention requires training on EMS-specific data, not general medical records.

The real-time analysis angle is interesting too. Instead of reviewing patient care reports after the fact, Amby can flag documentation gaps while the medic is still completing the report. That means fewer errors make it into the billing pipeline in the first place, which reduces denial rates downstream.

The Verdict

I think this is one of those applications where AI genuinely makes sense and isn’t just a buzzword stapled to a workflow tool. The manual review process in EMS billing is repetitive, rule-based, and error-prone. Those are exactly the conditions where AI performs well. And the economic incentive is clear: if an agency can reduce its billing staff from 12 people to 3 while improving claim accuracy, the ROI calculation is straightforward.

The risk is trust. Healthcare organizations are notoriously slow to adopt new technology, and EMS agencies are even more conservative than hospitals. These are organizations where a billing error doesn’t just cost money. It can trigger a federal audit. Convincing an agency director to let AI handle compliance-sensitive billing decisions is a sales challenge that pure product quality won’t solve alone. They’ll need case studies, pilot programs, and probably a few agencies willing to run Amby alongside their existing teams before going all-in.

There’s also the question of accuracy thresholds. In EMS billing, 95% accuracy isn’t good enough. A 5% error rate on 100 daily calls means 5 problematic claims per day, which is 150 per month, which is enough to attract payer scrutiny. The bar for automated billing in this space is essentially “as good as your best human reviewer, but faster.” That’s achievable with current AI, but it requires extremely clean training data and constant model updates as regulations change.

In 30 days, I’d want to see accuracy metrics from a live deployment. Not demo numbers. Real claims processed through real payers. At 60 days, the metric that matters is denial rate comparison: Amby-processed claims versus the agency’s historical baseline. By 90 days, if they can show a measurable increase in revenue per transport (through better coding) and a decrease in days-to-payment, that’s the kind of proof point that gets agency directors to pick up the phone.