Avelis Health Already Erased $300K in Bad Medical Bills, and They Are Just Getting Started

The Macro: Self-Insured Employers Are Bleeding Money on Claims Nobody Checks

Here is a number that should make every CFO lose sleep: self-insured employers in the United States waste more than $60 billion every year on medical claims that should never have been paid. Not fraud in the dramatic, someone-went-to-prison sense. Just billing errors, upcoding, duplicate charges, and services that were billed but never actually delivered.

About 65 percent of large employers in the US are self-insured, meaning they pay for employee healthcare directly rather than buying a policy from an insurance company. They use third-party administrators (TPAs) like Aetna or Cigna to process claims, but the TPA’s incentive is to process claims efficiently, not to scrutinize them. The faster claims get paid, the smoother the operation runs. Nobody is rewarded for slowing things down to check whether a $47,000 hospital bill is accurate.

The traditional approach to claims auditing is to hire a firm that samples a small percentage of claims and reviews them manually. They might look at 5 or 10 percent of total claims volume. The rest goes unchecked. The math is grim: if 3 to 7 percent of claims contain errors and you are only auditing 10 percent of volume, you are catching a fraction of the waste.

Companies like Cotiviti and Optum do claims auditing at scale, but they are massive, entrenched incumbents that primarily serve health plans rather than employers directly. Their technology stacks are aging and their pricing reflects their market position. Smaller players like ClaimDOC and ELAP Services focus on reference-based pricing rather than claims auditing per se.

The gap in the market is a modern, AI-native claims auditing platform that reviews 100 percent of claims in real time, catches errors before payment rather than after, and serves self-insured employers directly. That gap is exactly where Avelis Health is planting its flag.

The Micro: A Harvard PM With Three Knee Surgeries and a $250K Medical Bill Problem

Angel Onuoha, the CEO, has a personal connection to this problem that goes beyond market research. He has had three knee surgeries since his mid-teens and was billed over $250,000 for his care. That experience turned into obsession, and the obsession turned into a company. Before Avelis, he was a PM at a large tech company and led his first startup to 60-plus enterprise partnerships with major financial institutions. He is a Harvard alum and a three-time founder.

His co-founder and CTO, Ahmad Shehu, previously built healthcare data platforms for the Nigerian government that served over 20 million people. They have worked together for four years, which matters in a space where the technical and regulatory complexity would strain any new partnership.

The Avelis Health platform works in three stages. First, machine learning models review 100 percent of incoming claims, not a sample, flagging anomalies in billing codes, charge amounts, and service patterns. Second, voice AI agents automate the retrieval of medical records from providers, which is traditionally one of the most time-consuming steps in claims auditing. Third, LLMs perform clinical validation, cross-referencing the billed services against the actual medical records to verify that what was charged for actually happened.

The product integrates into existing employer and TPA systems in real time, meaning errors get caught before claims are paid rather than recovered after the fact. This is a meaningful distinction. Recovery audits are adversarial, slow, and recover pennies on the dollar. Prepayment audits prevent the waste from happening.

They came through Y Combinator’s Summer 2025 batch and have a team of three. Before pivoting to the employer-focused platform, they built a consumer tool that helped 60 patients reduce their medical bills by over $300,000 in under five months. That consumer tool validated the technology and the approach. The enterprise pivot is where the real money is.

The target savings for employers is 3 to 7 percent of annual claims spend. For a company with 5,000 employees spending $50 million a year on healthcare, that is $1.5 to $3.5 million in annual savings. The ROI case is trivial to make.

The Verdict

I think Avelis Health is attacking one of the most obvious and largest inefficiencies in American healthcare. The $60 billion waste number is not speculative. It is well-documented by industry analysts and acknowledged by the employers who are losing the money. The question has never been “is there a problem?” It has been “can technology actually fix it?”

The three-stage approach is smart. ML for pattern detection, voice AI for record retrieval, and LLMs for clinical validation. Each stage addresses a specific bottleneck in the traditional audit process, and together they enable 100 percent claims review, which is something no human audit team can deliver at a reasonable cost.

The risk is sales cycle length. Self-insured employers are risk-averse by nature. Convincing a Fortune 500 CFO to let a three-person startup audit their medical claims, even when the savings are obvious, requires trust, compliance approvals, and pilot programs. Cotiviti and Optum may be slower and more expensive, but nobody gets fired for hiring the incumbent.

Thirty days, I want to see a signed pilot with a self-insured employer of meaningful size. Sixty days, I want data on the actual error rate the platform catches versus what traditional audits find. If they are catching 3x more errors, the product sells itself. Ninety days, the question is whether the voice AI for medical record retrieval actually works at scale or whether it hits the wall that every healthcare automation product hits: providers who refuse to cooperate with automated systems. If the full pipeline works end to end, this company saves employers billions. That is not hyperbole. The math is that straightforward.