I paid for my medical care. Then I received $27,000 in bills for services I had already paid for.
When I spoke with the hospital on Oct. 6, I was given an explanation that should concern every American who has ever opened a medical bill: According to the hospital, an AI-enabled process within its Epic system identified that I had insurance and billed that insurance anyway, despite my having arranged and paid for the services as a self-pay patient.
The hospital has now corrected the disputed charges and credited my account. But the correction leaves a much bigger question unanswered.
How many people are paying bills they should never have received?
I am a father of four who has undergone two back surgeries and faces more treatment. I have enough experience navigating healthcare to know that a bill can be wrong. I also run a public relations business. I know how to reach leadership, escalate a complaint, and make clear that I am prepared to pursue public scrutiny and legal action.
It should not take those advantages to get an accurate bill.
Imagine a mother raising three children after losing her husband. She opens an envelope demanding thousands of dollars. She does not have a communications team, an attorney on speed dial, or hours to spend reconstructing transactions between a hospital and an insurer.
She has groceries to buy. Children to get to school. A job she cannot afford to miss.
Does she challenge the bill? Or does she assume the hospital must be right and start paying?
That hypothetical mother is the person Washington should have in mind when it examines the expanding use of artificial intelligence in medical billing.
The warning signs extend beyond my experience. Reuters reported on Sept. 24 that a Blue Cross Blue Shield Association study attributed nearly $1 billion in additional spending over two years to more intensive billing associated with AI tools. The association raised concerns about increases in documented conditions without corresponding increases in treatment. Those findings are not proof of fraud, but they demand scrutiny.
My experience involves a different question: how a paid self-pay arrangement became insurance billing and an enormous disputed balance. Both issues expose the need to examine what these systems are doing, who approves their actions, and who bears the consequences when they fail.
Epic, the company behind MyChart, publicly promotes revenue-cycle tools that verify insurance, support self-pay arrangements, and automate administrative work. Its own website presents these capabilities as ways to empower patients and maximize revenue.
Patients deserve to know what happens when those goals collide.
Finding an insurance policy should not make a patient’s payment disappear. An automated workflow should not erase a documented agreement without review. And a hospital should never send a demand for thousands of dollars without reconciling the money it has already received.
I cannot independently establish which software component caused my bills, whether it belonged to Epic or another vendor, or whether the hospital configured it incorrectly. That is precisely why an investigation is necessary. Patients should not have to reverse-engineer hospital software to understand what they owe.
The White House and Congress should investigate AI-driven medical billing, including Epic’s role in the episode the hospital described to me. They should demand the records that can establish what happened: payment histories, account changes, insurance submissions, software actions, and human approvals.
The investigation must also examine the hospitals deploying these systems and the insurers processing their claims. Responsibility cannot disappear into a chain of vendors.
I am calling for a suspension of AI-driven billing practices pending an independent review of their accuracy and patient protections. At minimum, systems that autonomously change a patient’s payment status, initiate claims, or generate payment demands should be paused until their operators can demonstrate that they reconcile prior payments and honor documented agreements.
Congress should require a named human reviewer for disputed bills, a clear explanation of every balance, and a halt to collection activity while an alleged duplicate charge is investigated.
Hospitals should also be required to review other accounts affected by a confirmed system error. Correcting the account of the patient who complains loudest is not enough. If a workflow produced one improper bill, leadership should determine whether it produced thousands.
I am building an AI company myself. I understand the promise of this technology. I also understand that deploying it creates an obligation to verify its work.
A system’s speed is no advantage when it accelerates mistakes and leaves patients to repair them.
Nor should “the AI did it” become an excuse for conduct that would be unacceptable if a person performed it. If an investigation establishes that an organization knowingly submitted false claims or pursued money it knew was not owed, authorities should investigate potential fraud. Software does not absolve the people responsible.
The national financial exposure could be enormous. But patients experience the damage one envelope at a time: a bill they cannot understand, a balance they cannot afford, and a fear that seeking medical care has endangered their family’s finances.
My disputed charges were corrected because I fought.
I’M A DOCTOR, AND I’M SICK OF WATCHING MY PATIENTS GO BANKRUPT JUST TO STAY ALIVE
The next patient may not know how. The next parent may not have the time. The next widow may pay because she is afraid.
Washington should act before an epic failure becomes an ordinary part of American healthcare.
Warren H. Cohn is CEO of RocketshipPR and a contributing columnist at JNS.