I’m an engineer, not a physician. For the better part of a year I worked on AI for primary care, and much of that time I spent close to primary care doctors and the people who keep their practices open. I went in assuming the hardest problems would be clinical. They weren’t. The thing I watched wear people down, week after week, was billing.

Not the medicine. The paperwork wrapped around the medicine.

A doctor sees a patient, does good work, documents it, and then the real ordeal begins. The claim goes out. Some weeks later a chunk of those claims come back denied, each for its own reason: a code that changed, a modifier a particular payer started requiring last quarter, a place-of-service mismatch, a prior authorization that was supposedly on file. Someone in the office, often the same person answering the phones, now has to decode each denial, fix it, and send it back. More than one in ten claims gets denied on first submission, and that number has been climbing. Every payer has its own rulebook, and the rulebooks keep changing.

For a small practice, this is not a nuisance. It is a tax on survival. The money is owed. It has been earned. But collecting it takes hours that a two- or three-person front office does not have, so a real share of it never gets collected at all. I sat with practices that were leaving meaningful revenue on the table not because they did anything wrong, but because chasing it would have meant ignoring the patients in the waiting room.

This is the part of the story that doesn’t make it into the headlines about consolidation. When we read that another independent practice sold to a hospital system or a private equity rollup, the easy explanation is that doctors wanted out, or that scale always wins. That is not what I saw. I saw physicians who loved running their own practice and were being pushed out of it by the back office. They didn’t want a new owner. They wanted to stop drowning in administrative work that has nothing to do with why they went into medicine. Selling was the exit from the paperwork, not from the work itself.

It gets worse when you look at care that’s specifically designed to help these patients. Medicare created a whole set of codes to pay practices for managing chronically ill patients between visits: checking in on the diabetic patient, coordinating the cardiology referral, reviewing medications for someone with five conditions. Chronic care management, remote monitoring, principal care management. The programs exist precisely because this work keeps people healthier and out of the hospital. But capturing them is so labor-intensive, with so much tracking and documentation, that most small practices can’t staff it. So the patients don’t get the care, and the practice doesn’t get the revenue that would have funded it. Everyone loses, quietly, by default.

Here is where I’ve landed, and why I’m writing this. The billing burden that’s pushing doctors to sell is, at its core, an information-processing problem. Decoding a denial, knowing this payer changed that rule, assembling the documentation a care-management code requires, this is exactly the kind of repetitive, rules-heavy work that modern AI has finally gotten good at. Not as a copilot that a tired staffer still has to babysit, but as something that can actually run the claims, work the denials, and surface the care a practice is eligible to deliver but doesn’t have the hands to capture.

I want to be careful here, because health care is full of software that promised to lift this load and instead added another login and another vendor to manage. The bar is not “we built an AI tool.” The bar is whether the work genuinely comes off the doctor’s plate. A human still has to own the parts where judgment and liability live: the coding decisions, the clinical calls, the final sign-off. But the bulk of the grind underneath that does not need a person doing it by hand anymore, and continuing to make people do it by hand is what’s breaking these practices.

What I believe, after watching this up close, is simple. A physician who wants to stay independent should be able to. That choice shouldn’t be decided by who has the bigger billing department. For most of the last decade, technology in this space has mostly served consolidation, making it easier for large systems to absorb small ones. It can just as easily run the other way. If the administrative weight that’s crushing independent practices can be lifted, the case for selling gets a lot weaker, and a lot of doctors who feel cornered right now would have a real choice again.

The patients in those practices are owed care they’re not getting. The doctors are owed money they’ve already earned. Both of those are fixable. We should fix them in a way that lets the doctor keep the practice.

Zack Spooner is an applied AI engineer.