AI won't fix America's healthcare system. It's helping people leave it. | Viewpoint

I was running a workshop last month when a physician leader asked me to predict what would happen next year. I don't make predictions. It's how you tell an academically trained futurist from someone who's adopted the title.

But her instinct was right. She was feeling the same thing many people in healthcare are sensing. Seismic shifts are happening that are going unaddressed. A year from now, I told her, many healthcare CEOs will discover that while they were automating workflows, patients and more than a few of their clinicians quietly found the exit.

The rush to automate is rational. Healthcare's economics don't work, workforce shortages are projected to deepen, and AI seems to have arrived just in time to deliver the efficiency the system has chased for decades. But every one of those automation decisions doubles down on a model built on three assumptions about American healthcare: enough workers to staff it, enough coverage to pay for it, and enough patience from patients to put up with it. That America no longer exists.

Today, one in five Americans lives in a primary-care shortage area. Since the enhanced ACA subsidies ended in December, premiums have jumped 58 percent, and enrollment has fallen by 2.6 million. Even in the biggest metros, where physician supply is richest, the wait for a new-patient appointment just hit a record high.

The system is broken, and everyone inside it feels it. But fixing it would require payers, providers, policymakers, and drugmakers to unite around a different economic model. No one is holding their breath. Instead, everyone retreats to their own small sphere of control.

Everyone, that is, except the people building the parallel system.

Many in the last wave of healthcare disruptors got entangled in the insurance process and died there. This new wave, however, is building outside of it, often through wellness and direct-to-consumer platforms that face fewer regulatory barriers. Wherever and however consumers define health, their door opens; they offer immediacy where the traditional system offers delay, and take cash where patients lack coverage. They’re AI-native by birth rather than retrofit.

Let's look at what's actually taking shape. Lotus AI has raised $41 million to scale free, 24/7 AI-powered primary care. Neko Health is entering the U.S. with 300,000 body-scan sign-ups and a $1.8 billion valuation. Eli Lilly sells Zepbound directly to self-pay patients. Hims & Hers built $2.35 billion in annual revenue selling online access to GLP-1s. In June, Midjourney, known for AI image generation, announced it was entering medicine through full-body ultrasounds. And by selling the results as wellness "body composition maps,” it sidesteps medical-device regulation while amassing the real product: the dataset.

The parallel system is removing the friction that the traditional system treats as immovable — long waits, missed workdays, gatekeeping, not feeling heard — and replacing it with control, data ownership, access, and a price up front rather than a surprise bill in the mail six weeks later. The appeal of the parallel front door is no mystery.

Patients aren't the only ones heading for the exit. The rate of physicians opting out of traditional Medicare has doubled since 2010 — not to retire, but to practice on their own terms: concierge medicine, direct primary care, cash pay. Others have left traditional practice altogether to review AI visits, prescribe through consumer platforms, and serve as medical directors for the companies building the parallel door. They didn't leave medicine. They left the way it's administered. The parallel system is recruiting from both sides of the exam-room door.

Research my foresight colleagues just completed on future health consumers points in the same direction. The deferential patient the system was built for is disappearing. Waiting is experienced as a malfunction, and managing one's healthcare has come to feel like a second job that people are “quiet quitting.” And even those who aren’t leaving the system may soon enough not be reaching for a door at all; they’ll send AI agents to compare, book, and negotiate. The first door those agents knock on won't necessarily be yours.

Skeptics make good points. Much of this next wave of disruption is questionable medicine that increases incidental findings and overdiagnosis, unvetted AI advice, or wellness theater. The two-tier system worry is real, too, though those tiers already existed, and a free AI front door could widen access for those who are priced out. But while traditional healthcare is busy winning this evidence debate, it’s losing the relationship.

The frontier AI companies have already figured this out. OpenAI has led with an aggressive consumer play, putting health guidance directly into millions of pockets and competing to capture the first interaction. More than 300 million people now bring health questions to ChatGPT each week, and U.S. users can connect their medical records. Then, in September, OpenAI announced a read-only Epic integration that brings authorized patient context into ChatGPT. Anthropic started on the institutional side—building tools for providers and payers while also enabling individuals to connect their health records to Claude. They started in different places, but both are heading to the same destination. By positioning themselves between patients and the system they own the relationship—and the traditional system is opening its own front door to let them in.

What the parallel system is really competing for isn't visits; it's the relationship, the trusted first interaction with the power to route everything downstream. One day, in the not-too-distant future, leaders of the traditional system may wake up and realize most of what they have left is serving the sickest and costliest while the trust, the data, and the margin live elsewhere.

Every decision to wire a workflow with AI is an assumption with concrete poured over it, on a payback schedule that matures in three to five years, right as the parallel front door reaches scale.

The question is not whether traditional healthcare should automate. It's whether their concrete assumes a patient who will still be walking through the door. The same technology that speeds up denials can also be deployed to answer a patient’s question at midnight, turn a three-month wait into a three-minute one, and show the price up front.

The front door is where relationships begin. The parallel system already knows this. And it's open all night.

Denise Worrell is the managing principal of Langrand’s Transformation and AI strategy practice.