Home-health operators are understandably cautious about putting AI in front of patients.

They worry about what patients will think. They worry about what families will think. And perhaps most importantly, they worry about what referral sources will think.

That concern is fair. Healthcare is built on trust and relationships. Technology should serve those relationships, and we should hold it accountable when it does not.

But I think we may be underestimating the risk on the other side:

The communication experience we have already accepted as normal.

A patient is discharged from the hospital and referred to home health, entering a system they are often completely unfamiliar with.

Who is coming to my house? When will they arrive? What happens next? Who do I call if I have a question?

Yet it can take days before an intake team successfully reaches that patient. Plenty of effort goes into it, but between wrong numbers, voicemails and someone else answering the phone, actually reaching the patient is hard.

Then, once care has started, the patient calls the agency and is asked to navigate an IVR system or wait in a queue that, during peak hours, can easily exceed ten minutes.

We have somehow normalized patients sitting at home listening to hold music while trying to get basic information about their care.

We scrutinize every new tool that touches a patient. We have never scrutinized the status quo of endless phone queues with the same intensity.

We Have Made Communication Too Reactive

CareBestie product screenshot of why people are calling and who they are

This is not because agencies don't care.

It is because their teams are overwhelmed.

Staff are answering phones, returning calls, tracking down clinicians, responding to questions and constantly putting out fires. There simply isn't enough capacity to communicate proactively with every patient.

So most communication happens after something has already happened.

The patient calls. The caregiver complains. The clinician gets a message. A poor survey comes back weeks later.

Technology gives us the opportunity to reverse that model.

Instead of waiting for patients to contact the agency, agencies can proactively check in between visits:

How are you doing?

Did your clinician arrive?

Do you understand what happens next?

Do you have questions about your medications?

Do you feel heard?

Is there anything you want your care team to know?

And they can do it consistently, 24/7 and in the patient's preferred language.

That is where I believe AI becomes genuinely valuable.

The Goal Is Not Less Human Interaction. It Is More.

Technology should handle the communication that otherwise would not happen and surface the moments where a person is actually needed.

A patient has a new concern. A caregiver is frustrated. Someone didn't understand the medication instructions. A patient feels their condition is getting worse. Someone simply wants to speak with a person.

Those are the moments the care team needs to know about. That is what we built CareBestie to find.

Michael is not making clinical decisions. His job is to understand what is happening, identify when human attention is needed and get the right person involved quickly.

We see this clearly in after-hours operations.

At one agency, roughly half of after-hours calls were non-urgent and could be handled without waking anyone. The other half required a person.

And the split reflected what patients were actually calling about: more than 80% of clinical-support calls were escalated to a human, compared with approximately 40% of scheduling calls.

When a person was needed, one showed up. On-call staff answered 92% of escalated calls, on average in just over a minute.

One nurse using CareBestie described an evolution that I think captures this well.

Initially, the technology could feel like another system evaluating the clinician. Over time, she began to see it differently: as another layer of communication that could identify patient questions or concerns between visits and give the clinical team an opportunity to follow up.

That is exactly the role technology should play.

Not replacing the nurse.

Not replacing the relationship.

Helping the agency hear what is happening during all the hours when nobody from the care team is physically in the room.

The Patients May Be More Ready Than We Think

Our oldest patient to have a conversation with Michael, CareBestie's AI agent, is 104 years old.

She spoke with him for six minutes about her care.

And she is not an isolated example.

In our July patient data, engagement was not concentrated among younger patients. Patients aged 65–74 and 75–84 had the highest combined completion rates, at 67.4% and 65.8%. Patients aged 85–94 still completed conversations at a 62.7% rate.

Once patients connected, completed conversations averaged just over three minutes.

One patient gave us perhaps my favorite reaction:

“If that was AI, it was the best one I ever heard in my life. They paid dearly for that, because I couldn't tell the difference between a real person.”

Of course, not every patient wants to speak with AI.

That is fine.

If a patient wants a person, the system should recognize that and connect them with one.

The objective should never be to force technology onto patients.

The objective is to make sure they are heard.

Only Then Can We Start Talking Seriously About Patient Experience

Before asking why patients aren't filling out surveys or why HHCAHPS scores aren't improving, we should ask a more basic question:

How frequently are we actually communicating with our patients?

If patients feel uninformed throughout the episode, calling them at the end and asking whether they were satisfied is probably too late.

But if an agency is checking in consistently, identifying concerns early, reinforcing education and closing the loop when something goes wrong, then we can start meaningfully improving the patient experience.

A home-health clinician might spend one or two hours with a patient every few days.

There are 168 hours in a week.

The enormous opportunity is what happens during the rest of them.

AI does not allow an agency to provide clinical care for all 168 hours.

But it can extend the agency's ability to listen, communicate and respond far beyond the physical visit.

And when something concerning surfaces, it can make sure the right human knows about it.

That is not replacing human connection.

It is creating dramatically more opportunities for it.

So perhaps the question for healthcare leaders shouldn't be:

“Are we comfortable letting AI talk to our patients?”

Perhaps it should be:

“Are we comfortable with how little we're talking to them today?”