How AI and automation are changing healthcare

My first real look at healthcare was not a hospital. It was a benefits administration office in South Florida, where I built the software single-handedly and learned how the system actually moves. It moves on paper. Faxes, forms, phone calls, prior authorizations, claims that bounce for a missing code and come back three weeks later. The care is modern. The plumbing is from another century.

That gap is where automation lives, and it is why I am more optimistic about healthcare than about almost anything else on this list. The upside is not a robot surgeon. It is a hundred million hours a year of administrative work that nobody will miss.

The boring, enormous wins

• Documentation. A visit that used to end with twenty minutes of typing now ends with a note the doctor reads and signs. The single biggest cause of burnout in medicine is the chart, and the chart is being written by a machine that listened.
• Prior authorization and claims. The back-and-forth between a clinic and a payer is a conversation between two sets of rules. Models can read both, assemble the evidence, and send a request that is right the first time. The appeal that used to take a week takes an hour.
• Scheduling and follow-up. No-shows, unfilled slots, reminders, refills, and the hundred small handoffs that fall through. This is the same work I automate for any business. Healthcare just has more of it.

The clinical wins, carefully

Imaging is the clearest case: a second reader that never gets tired, that flags the scan a radiologist should look at first. Triage is close behind. So is the quiet work of watching a patient's numbers between visits and noticing the drift before it becomes an emergency. I wear a tracker. The bed behind my desk, in the 3D version of this page, is where those numbers live. Imagine that for a hundred million people, with a clinician in the loop.

The word that matters is assist. Every one of these works best when a person makes the call and the machine makes the call easier to make well. I am not interested in automating a diagnosis. I am interested in automating the four hours of work that stood between a doctor and the patient.

What I think about the mental health side

I now spend my days at a mental health organization built to reach more than a billion people, and the only way to reach a billion people is automation. Not a chatbot therapist. The intake, the matching, the scheduling, the follow-up, the paperwork, the measurement of whether anything is working, all of it done by software so that the human hours go where only a human will do. The scarce thing in mental health is a person who listens. Everything else can be built.

The line I would not cross

Automation that makes care cheaper by making it worse is not automation. It is a cut with better branding. The test I apply is simple: would I want this for my own kids? The answer decides what gets built.