American healthcare is broken in a way you can measure

American healthcare is broken. Not in the vague, everyone-nods way, but in a specific, mechanical way you can point to. On ACA marketplace plans, roughly one in five in-network claims is denied.¹ Fewer than 1% of those denials are ever appealed. Of the few that are, about a third are overturned.

Sit with those three numbers for a moment. Care is denied at industrial scale. Almost nobody pushes back. And when somebody does, they win a third of the time. That means an enormous amount of legitimate, medically necessary, ultimately payable care is forfeited every year, not because the clinical case was weak, but because nobody had the hours to make it.

That is not a market working. That is a war of attrition that one side stopped showing up for.

We are a medical student and a computer scientist, and this is the absurdity that sits between our two fields: a physician’s judgment, formed by years of training and an actual examination of an actual patient, routinely overruled by a form letter from someone who has never seen that patient.

We didn’t trust our outrage, so we tested it. We made more than a thousand phone calls: to oncologists, practice administrators, billers, revenue-cycle leads, prior-auth staff, people who used to work inside payers. We asked the same question every time: where does the fight actually break down?

The answer was not the one we expected. The fight is not lost on merits. It is lost on logistics.

Practices don’t lack arguments; they lack hours. Prior authorization alone consumes about thirteen hours of physician and staff time every week.² An appeal is a small research project on top of that: find the payer’s policy, pull the chart, match evidence to criteria, draft the letter, get the physician’s signature, submit through the right channel, hit the deadline, chase the response. All of it done by the busiest people in the building, on top of the jobs they were hired to do. So the rational move, case after case, is to eat the denial and move on. Multiply that quiet decision across every practice in the country and you get the 1% appeal rate.

Once you see the problem as logistics, the solution stops being mysterious. Nobody needs a smarter argument; the existing arguments already win a third of the time on the rare occasions anyone makes them. What is missing is the machinery to make them every time.

So that is the first thing we built. Inveto reads the denial and the clinical record, drafts the prior-auth response or the appeal with every material fact bound to a source document (or flagged, never invented), and puts it in front of the physician to review and attest. Then it does the unglamorous part: submission, deadlines, the payer’s response, the outcome, the remittance. “Approved” and “paid” are two different events in this system, and the gap between them is where practices quietly get shorted.

The obvious response to denials, and the one a wave of new AI tools has taken, is to write the letter faster. But the letter was never the bottleneck, and a plausible letter that invents a fact is worse than no letter at all. The bottleneck is the operation: the tracking, the deadlines, the follow-through, the proof. That is what has to be automated, so that is what we automated.

Something else happens when you run that operation, and it is the reason this company exists.

Every contested case answers a question nobody has been recording: what actually works? Which evidence, against which payer, under which policy, changed the outcome, and which didn’t?

Consider where health data lives today. Medical-records systems record the care. Clearinghouses record the claims. Payers keep their decision logic to themselves. Nobody holds the link between the argument made and the outcome it produced. And since 99% of denials are never appealed, that record is not merely hidden. It has never existed.

Every case that moves through Inveto creates it: a de-identified, outcome-linked record of a real coverage decision, with the payer, the policy, the denial reason, the evidence, the attestation, the outcome, and the dollars. Reuse rights are written plainly into our contracts, and benchmarks flow back to the practices whose work generates the record.

Healthcare data efforts usually die at the source: the data must be bought, scraped, or begged for. This record is different. It is the byproduct of work practices urgently want done anyway. The workflow funds the record; the record compounds with every case. We call it the coverage-decision graph, and we think it becomes the definitive account of what health-insurance decisions actually do in the real world.

Follow the graph forward and it reaches a second, larger moment: the one where insurance gets chosen.

Americans pick health plans on two numbers, premium and deductible, and both are approximately fiction. The premium is what you pay while you are healthy; the deductible gestures at what you will pay when you are not. Neither answers the only two questions that matter: what will this plan actually cost me over a year of my real life, and will it pay when I need it?

The first question is arithmetic that nobody does. Surely does it: a deterministic twelve-month simulation of what a plan will truly cost out of pocket, from premiums net of subsidies through deductibles, drug tiers, coinsurance, networks, and out-of-pocket maximums, run against your actual conditions and medications rather than an actuarial average.

The second question, whether it will actually pay, is one no comparison tool can answer today, because answering it requires a record of how plans behave when their decisions are contested. That record is the graph. Every denial fought on one side makes the answer sharper on the other.

That is the whole design. One product fights the system’s worst failure and gets paid for doing it. The fight produces the record. The record makes the next fight smarter, and the next enrollment decision with it. It compounds, and it has already started.

A thousand phone calls taught us that everyone inside this system knows it is broken, and that nearly everyone has made a kind of peace with it. We haven’t. The system depends on people not fighting. We intend to change the math on that.