AI Is Scaling Healthcare Costs Because the System Was Built That Way

In the employer benefits work I do, I watch plan sponsors absorb cost increases they cannot fully explain and deploy defenses they cannot fully trust. The billing is more sophisticated than it was five years ago. The defenses have not kept pace. What has changed is not the system. It is the speed at which the system operates.

AI is not changing healthcare. It is scaling the economic logic that already exists within it. Across the country, health systems are deploying AI-powered revenue cycle tools designed to maximize billable revenue from every patient encounter. The market for these tools has already surpassed $20 billion and is growing rapidly, driven by one clear objective: capture more from the same clinical interaction.This is not innovation in care delivery.  It is optimization of reimbursement.

And as that optimization accelerates, the downstream effects are beginning to surface. Stop-loss premiums rose 9.4% in 2024 among tracked health plans, with employers maintaining comparable coverage seeing increases closer to 11.5%. Claims exceeding $1 million per million covered employees jumped 29% year over year. Cost is not just growing. It is concentrating. The industry is calling this an AI problem. It is not.

We are defending against a system designed to produce the problem

The current response has been to build better defenses. Employers are layering in payment integrity vendors to audit claims after the fact. They are shifting to reference-based pricing models to negotiate against inflated charges. They are relying on third-party administrators to create more transparency into a system that was never designed to be transparent.

Each of these strategies can reduce cost at the margin. None of them change how the cost is created. They are all operating at the same point in the chain: after the claim has been generated, coded, and submitted. That is not control. That is containment.

And containment, no matter how sophisticated, is still downstream of a system that rewards billing intensity as its primary economic output. You cannot out-defend a system you have not redesigned.

The economic loop AI is accelerating

For decades, healthcare has operated within a simple and largely unquestioned cycle. Clinical documentation determines reimbursement. Reimbursement models reward greater intensity. Greater intensity increases total cost.

That cost is then distributed across employers, employees, and public programs, where it becomes a budget problem rather than a design problem. AI does not disrupt this loop. It compresses it.

What once required manual review and coding judgment can now be executed at scale, in real time, across millions of encounters. The result is not a new system, but a faster version of the existing one: more precise, more consistent, and more effective at extracting revenue from the same underlying clinical activity. Efficiency, in this context, does not create value.It amplifies misalignment.

Why procurement-led solutions are hitting a ceiling

In response, a growing number of employers have moved to more advanced procurement strategies. They are working with independent third-party administrators, implementing reference-based pricing, and restructuring benefits to reduce reliance on traditional carrier networks. Some have gone further, contracting directly with primary care providers or building employer-sponsored clinics to regain a degree of control over access and cost.

These models have produced measurable outcomes. A Milliman actuarial study found that employees enrolled in a direct primary care option reduced overall healthcare demand by nearly 13% and cut emergency department use by more than 40% compared to traditional plan peers. That is not marginal. That is structural deflection of cost before it enters the billing system.

But even these approaches face a ceiling, and the stop-loss market is where it becomes visible. Stop-loss premiums rose 9.4% in 2024 among tracked health plans, with employers maintaining comparable coverage seeing increases closer to 11.5%. Claims exceeding $1 million per million covered employees jumped 29% year over year, a signal that cost is not just growing, it is concentrating. Employers are absorbing more risk precisely because the underlying delivery system keeps generating it.

The problem is not that these strategies are ineffective. It is that they are incomplete. The savings represent what is achievable within a system you do not control, which makes the distance to full redesign even more visible.

The shift from procurement to system design

The next phase of healthcare transformation will not be defined by better negotiation. It will be defined by where control is established.

Employers and operators who are moving upstream are doing something fundamentally different. They are not simply purchasing care more effectively. They are restructuring how care is accessed, delivered, and managed.

In direct primary care models, payment is decoupled from volume entirely, removing the incentive to increase billing intensity at the front door. That structural shift matters, but only when DPC is designed as a layer within a broader benefits architecture that includes catastrophic or high-deductible coverage, not as a standalone displacement of comprehensive insurance. When employer groups build DPC into their plan design intentionally, controlling pricing logic and care coordination, they change where cost originates. The AI-driven revenue cycle management market now exceeds $20 billion and is projected to nearly triple by 2030, a trajectory that makes the contrast with volume-decoupled models sharper, not softer. In employer-aligned networks and value-based arrangements, downstream referrals and utilization are actively managed, not passively received. In employer-sponsored clinics, the point of entry itself is redesigned to prioritize continuity, prevention, and cost control before high-cost services are ever introduced.

These are not procurement strategies. They are operating models. And they change where cost originates, not just how much of it gets recovered afterward.

The verdict

AI is not driving healthcare cost inflation. It is exposing a system that has always been designed to reward it, and making that system more efficient. As long as reimbursement is tied to intensity, every technological advancement will move in the same direction: toward maximizing it. A benefits leader sitting in a plan renewal meeting today is not facing an AI problem. They are facing a design problem that AI has made impossible to ignore.

The math of staying downstream is no longer abstract. Average employer-sponsored family coverage now runs nearly $27,000 annually, and total health benefit cost per employee is projected to rise 5.8% in 2025 even after cost-reduction measures, marking the third consecutive year of increases above 5%. That is not a trend line. That is a compounding consequence. The employers I work with who remain in a reactive posture face the same pattern: stop-loss renewals that exceed projections two years running, plan designs that erode because cost trend is outpacing contribution increases, and carriers repricing or walking away from self-funded accounts entirely. Each of those outcomes is the predictable result of defending against a system rather than redesigning your position within it. The question at every renewal is not whether costs will rise. It is whether you are willing to keep absorbing increases that a different structural approach would have prevented.

The question is no longer whether we can detect or negotiate inflated costs. It is whether we are willing to operate within a model where those costs can be generated in the first place. AI is not the disruption. It is the mirror.

Photo: Meriel Jane Waissman, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.