It's Not What You Do. It's How. Why I Built AI Into a Healthcare Platform and Why I Almost Didn't

There's a conversation happening right now about AI in healthcare. Most of it is loud. Most of it is scared. And honestly, most of it is missing the point.

People aren't afraid of AI. They're afraid of what happens when AI is used irresponsibly. In healthcare, that fear isn't paranoia. It's justified.

Because we've already seen what happens.

Insurance companies deploying AI tools that deny Medicare claims at scale, with error rates so high that federal judges called them out in court. Patients waiting for surgeries that keep getting pushed back. Elderly nursing home residents losing therapy coverage. Families fighting systems they don't understand for care their loved ones were already entitled to.

That's not a technology problem. That's a responsibility problem. And the difference matters enormously.

10,000 Calls

I've taken over 10,000 Medicare Advantage calls. I know what it sounds like when a patient finds out their procedure was denied. I know the confusion. The fear. The exhaustion of fighting a system that feels designed to wear you down.

I also know something else most people don't talk about. Prior authorizations exist for a reason. They were designed to protect patients - from unnecessary procedures, from bad actors, from a system that without checks and balances gets taken advantage of from every direction. Patients. Providers. Payers. Everyone has a point of view. Everyone has a grievance. And everyone is right about some of it.

The problem was never the idea behind prior authorizations. The problem is that nobody built a system that made them work the way they were supposed to. So I built one.

Why I Almost Stopped

AI in healthcare terrifies me. Not because AI isn't powerful. Because it is. And in healthcare, where the data is the most sensitive, the stakes are the highest, and the consequences of getting it wrong are measured in human suffering, power without responsibility isn't innovation. It's liability.

Two things kept me up at night about AI in healthcare: PHI exposure and hallucinations.

Patient health information is the most protected data in existence for good reason. One breach. One exposed record. One subpoena. The damage isn't just legal. It's human.

And hallucinations - where AI confidently generates wrong information - in a billing workflow doesn't mean a typo. It could mean a wrong appeal argument. A misquoted guideline. A denied claim that should have been approved.

I wasn't willing to build something that created those risks. So I engineered around them entirely.

How We Built Around the Risk

And hallucinations? There is always a human in the loop. Always.

The AI writes the appeal argument based on CMS guidelines. A human connects it to the patient record and makes the final call. The AI handles pattern recognition. The human handles judgment. Because judgment isn't something you automate in healthcare. It's something you protect.

What Responsible AI Actually Looks Like

This is what responsible AI actually looks like. Not a disclaimer at the bottom of a webpage. Not a compliance checkbox. Not a promise that the model is probably fine.

Architecture that makes harm structurally impossible. And a human who is always accountable for the final decision.

AI Has Always Been a Change in Times

AI is not the enemy. I told my sister this recently when she mentioned getting pushback about AI adoption in her HR firm. I told her to show Hidden Figures at her next meeting.

Those women were human computers. Then the machines came. And the ones who thrived weren't the ones who fought the technology. They were the ones who learned to work alongside it. Who brought their judgment, their experience, and their humanity to a process the machine alone could never get right.

AI has always been a change in times and job descriptions. The question was never whether to use it. The question is whether we use it responsibly.

In healthcare, where the data is sacred and the stakes are human, the answer to that question isn't optional. It's everything.

That's why I built Pre Auth Health the way I did. Not just to fix a broken prior authorization process. But to show that AI in healthcare can be done right. PHI-free. Human supervised. Accountable at every step.

Because the goal was never to replace human judgment. It was to protect it.