AI in Healthcare Has Graduated. The Hard Part Starts Now.

VectorCare recently published a perspective in Becker's Hospital Review on where healthcare AI goes after the hype cycle. The short version: the industry has solved the easy problem. The hard one is just getting started.

The Documentation Wave Is Over

The numbers are no longer debatable. Cleveland Clinic has documented over a million patient encounters using ambient AI scribes. Mass General Brigham cut after-hours charting from 90 minutes to 30. CommonSpirit reports $100 million in annual AI savings across 230 deployed solutions. Healthcare AI has graduated from pilot programs to enterprise deployment — at least for clinical documentation.

But documentation was always the most tractable problem. The input is structured speech, the output is a structured note, and the feedback loop is immediate. It was the right place to start. It is not the place to stop.

The 97% Problem

One executive at the Becker's CEO + CFO Roundtable framed it this way: health systems need to stop thinking about "patients" and start thinking about the 97% of people who don't wake up thinking of themselves that way. That reframing exposes how much of the care delivery chain sits outside the clinical encounter.

A patient doesn't experience healthcare as a series of clinical visits. They experience it as a series of logistics problems — getting to the appointment, receiving the equipment they were prescribed, understanding what happens next, and making it to the follow-up. The encounter is 15 minutes. The logistics surrounding it can take days. When the logistics break down, the care plan breaks down with them.

Why Operational AI Is the Next Frontier

Three forces are converging to make this urgent. The outpatient pivot is accelerating — Ascension is contracting from 139 to 91 hospitals while Sutter approved $450 million for 50 new ambulatory surgery centers. Every site added to a distributed network creates exponentially more logistics connections to manage. Value-based care models mean a missed follow-up or undelivered piece of DME is not just a service gap but a financial exposure. And with nearly $1 trillion in Medicaid cuts projected over the next decade, health systems need to do more with significantly less.

This is where patient logistics automation creates outsized value. Not by replacing care coordinators, but by handling the scheduling, matching, routing, and exception management that consume their day — so they can focus on patients who need human judgment.

Infrastructure, Not Applications

The reason AI scribes deployed so quickly is that they plug directly into the clinician's existing workflow. No new system to learn, no context switching. The same principle applies to operational AI: it has to be embedded in the EHR, not bolted on as a separate portal. FHIR APIs make this possible — when a logistics platform reads discharge orders, patient demographics, and insurance information directly from the EHR, the coordination workflow starts automatically.

This is the architecture VectorCare has been building: FHIR-native patient logistics infrastructure that connects transportation, home health, DME, and last-mile coordination directly into clinical workflows. The care team stays in their workflow. The logistics happen in the background.

Read the Full Perspective

The full article in Becker's Hospital Review covers the data, the market forces, and the integration architecture in detail. Read it here.

The health systems that figure out operational AI will be the ones that can actually deliver on the distributed care model they are investing billions to build. The ones that don't will have a beautiful network of facilities and an AI-powered EHR — and patients who still can't get there.