There's a version of "AI in healthcare" that I don't believe in: the one where a chatbot replaces the person who calls to check on your mother. That's not what we're building at CareTrack, and it's not what this article is about.
What I do believe in, because I watch it work every day, is AI applied to the operations of care management: the routing, the prioritizing, the documenting, the pattern-spotting. The unglamorous machinery that decides whether a care team spends its day on the patients who need it most or on the busywork around them. When that machinery gets smarter, three things happen. Our team gets more efficient, patients get better care, and providers get more visibility into what's happening with their panel. Here's how those connect.
The problem AI is actually good at
CareTrack's care team handles an enormous volume of between-visit work. At our largest partner practice alone, that's tens of thousands of live patient calls a year, plus the readings streaming in from remote patient monitoring devices every morning.
The hard question in that volume has never been "can we do the work?" It's "what should we do first?" Which of this morning's blood pressure readings represents a trend that's genuinely turning, versus a skipped medication, versus a cuff worn over a sweater? Which of today's scheduled chronic care management calls should move to the top of the list because something in the patient's recent readings changed?
That's a prioritization problem across thousands of data points, and it's exactly the kind of problem AI is good at. Not replacing the judgment of a care coordinator, but making sure their judgment is pointed at the right patient at the right moment.
Efficiency that patients can feel
When people hear "operational efficiency," they picture cost-cutting. What it actually looks like in our world:
- Faster response to the readings that matter. Smarter triage means the out-of-range reading that needs a phone call gets one sooner, because it isn't waiting in line behind twenty readings that didn't.
- More prepared conversations. When a coordinator picks up the phone, the relevant history (recent readings, last month's notes, open follow-ups) is assembled in front of them instead of hunted for. The call time goes to the patient, not the chart.
- Less documentation drag. Every interaction we have lands in the provider's EHR. Assembling that documentation used to compete with actually caring for people. Increasingly, the drafting and structuring is assisted, and the human reviews and signs off.
None of that is a robot delivering care. All of it is the same care team, with more of their hours pointed at patients. The measure I care about isn't "calls handled per coordinator." It's how quickly a worrying trend turns into a human conversation, and how much of every call is spent on the patient instead of the paperwork.
Better care is the point, not the byproduct
The clinical logic of everything we do rests on one idea: problems caught between visits are cheaper, safer, and easier to fix than problems caught in the ER. A patient whose blood pressure has been drifting up for two weeks is a medication adjustment. The same patient six weeks later can be a stroke.
AI widens the net for catching those drifts. A human reviewing this morning's readings sees today. A model watching the same data can flag the slow, unremarkable-looking slide that never trips a single-day threshold, and put it in front of a coordinator with the question, "does this patient need a call?" The escalation rules stay exactly where they belong: defined by the patient's physician, executed by our team, documented in the chart.
Transparency providers can audit
Here's the part I think gets talked about least. Providers who outsource care management have historically bought a black box. A vendor says the calls happened, the invoice arrives, and the practice hopes the two are related.
Our answer to that has always been EHR integration: every touchpoint, reading, and minute of time documented in the practice's own system, where the provider can see it. AI raises what that documentation can do. Structured, consistent records of every interaction mean a practice doesn't just get more data, it gets answerable data: which patients are engaging, where readings are trending, what the care team did about it, and when. When your quality team asks how the hypertension population is doing this quarter, the answer comes from the chart, not from a vendor's slide deck.
Transparency is also the standard we hold ourselves to on AI specifically. Anything that touches patient care runs under the same rules as the rest of our operation: physician-defined protocols, human review, and a documented trail. If a tool can't meet that bar, it doesn't ship.
Where this goes
The next few years of AI in care management will produce a lot of demos and a lot of noise. Our filter for what's real is simple, and it's the same filter we'd suggest any practice apply to any vendor, including us:
- Does it put more human attention on patients, or less?
- Does the physician still define the rules?
- Can you see the work in your own EHR?
If the answer to all three is yes, AI makes care management better. If not, it's just automation wearing a stethoscope.
Want to see what this looks like on your own patient panel? Book a 20-minute walkthrough.