This Isn’t an AI Prediction. This Is AI at Work.

154,552 outreach actions. 48,352 AI calls. 174.2 hours of patient conversations. Zero staff involvement. Patient-responsibility revenue doubled within 30 days.

There is no shortage of conversation right now about what artificial intelligence might eventually do for healthcare.

I’m less interested in what AI might do.

I want to know what it is actually doing.

As Co-Founder of Reclaimly, I’m obviously close to this technology. But I’ve also spent decades working around patient responsibility, billing and collections.

I’ve watched healthcare technology evolve. I’ve watched new platforms promise greater efficiency, better collections and fewer manual processes.

And one lesson has remained pretty consistent:

Activity is not a result. Technology is valuable when it changes the outcome.

So rather than talk about what autonomous AI might do someday, I want to share what we’re seeing from one practice using Reclaimly today.

The headline result is difficult to ignore:

Patient-responsibility revenue doubled within 30 days after deployment.

And the outreach behind that result was executed with zero staff involvement.

That’s where this story gets interesting.

What Did the AI Actually Do?

Let’s start with the work.

For this one practice, Reclaimly executed:

154,552 total outreach actions

Including:

67,107 texts sent

38,924 emails sent

48,352 AI calls placed

183 letters mailed

The communication mix was:

43% text

28% email

29% AI calls

These aren’t projections or estimates of what AI could potentially accomplish.

They represent actual outreach executed for one practice.

That distinction matters because the real story isn’t simply that AI was present in the workflow.

The AI was doing the work.

48,352 AI Calls Make the Shift Tangible

The calling data provides an even clearer picture of what that means operationally.

The AI placed:

48,352 calls

Of those:

16,203 were answered

That represents a 33.5% answer rate.

Another 16,225 voicemails were left.

The average answered call lasted 39 seconds.

Collectively, those conversations produced:

174.2 hours of patient talk time

And 1,892 patients called in.

Think about what that represents operationally.

Tens of thousands of calls. Thousands of answered conversations. Thousands of voicemails. More than 174 hours of patient talk time.

And none of that outreach became another calling list for the practice’s staff.

The Two Numbers That Matter Most: ZERO and 2X

The 154,552 outreach actions demonstrate scale.

But after spending decades around patient responsibility, I don’t believe that’s the most important number in this story.

I think two numbers matter more:

ZERO

and

2X

Zero staff involvement in performing the outreach.

2X patient-responsibility revenue within 30 days after deployment.

That’s the connection that matters.

The activity tells us how much work was executed.

The revenue result tells us why that work mattered.

We’ve Had Automation for Years. This Is Different.

Healthcare doesn’t have an automation shortage.

We have software, portals, dashboards, work queues, alerts and reports.

Much of that technology has improved healthcare operations.

But there is an important difference between technology that helps someone perform work and technology that performs the work itself.

A dashboard still needs someone to look at it.

A queue still needs someone to work it.

An alert still needs someone to respond.

A list of patient balances still needs someone to act on those balances.

That has traditionally put the human employee directly in the middle of the workflow.

AI gives us an opportunity to rethink that model.

Instead of only asking:

How can we give staff better tools?

We can begin asking:

Why does staff need to perform this particular work at all?

That is a very different conversation.

Don’t Add AI to a Broken Workflow

This is where I believe organizations need to think differently about AI.

If we start with the existing workflow and simply look for places to insert AI, we may miss the larger opportunity.

Start with the outcome.

Then work backward.

What has to happen to get there?

Which steps genuinely require human judgment?

Which steps exist because historically there was no other way to perform them?

And which can now be executed autonomously?

That’s where workflow redesign begins.

I’ve used this line before because I think it captures the shift:

Once the SOP is rewritten, the agent stops being a demo and starts being a worker.

That’s very different from adding another technology layer to an existing process.

You’re rethinking how the work itself gets done.

From Software → Staff → Action to Signal → AI → Action

For most of my career, healthcare technology has followed a pretty familiar model:

Software → Staff → Action

The software provides information.

Someone interprets it.

Someone takes action.

Someone follows up.

Someone monitors what happens next.

Autonomous AI introduces a different possibility:

Signal → AI → Action

The technology isn’t simply telling someone what to do next.

It can execute the work.

That’s why I don’t think the real opportunity for AI in the revenue cycle is simply greater employee productivity.

I believe the bigger opportunity is workflow redesign.

Why Patient Responsibility Is a Natural Place for This Shift

Patient-responsibility recovery has always involved a tremendous amount of repetitive operational work.

Once a patient balance exists, execution matters.

Outreach has to begin. Communication has to continue. Different channels may need to be used. Patients may have questions. Follow-up cannot simply stop because staff members are busy with other responsibilities.

That’s the operational reality I’ve seen throughout my career.

The problem isn’t usually knowing that a patient balance needs attention.

The challenge is consistently executing all the work required to move that balance toward resolution.

Historically, the capacity to do that work has been closely tied to staff capacity.

AI changes that equation.

And that is what I find so significant about the results from this practice.

How Should We Measure AI in Healthcare?

There’s going to be an enormous amount of discussion about AI in healthcare.

There will be new platforms, new features, new demonstrations and plenty of predictions.

Some will ultimately change healthcare operations.

Others won’t.

I think operators need a simple way to cut through that noise.

Instead of beginning with:

“Does it use AI?”

Ask:

What work is it actually doing?

How much work is it doing?

How much staff involvement does it require?

What measurable outcome changed because of it?

Those are the questions that interest me.

Because ultimately, healthcare organizations don’t need AI for the sake of having AI.

They need better outcomes.

From Software That Helps People Work to Software That Works

After decades around patient responsibility, this is the change I find most significant.

For most of my career, healthcare technology has been built around people operating software.

The software changed.

The screens improved.

The data got better.

Processes became more digital.

But ultimately, a person remained somewhere inside the workflow, performing the work.

AI gives us the opportunity to reconsider that assumption.

Not everywhere.

Not for every task.

And certainly not where human judgment and interaction are essential.

But for highly repetitive operational work, we can begin asking a fundamentally different question.

Not:

How can technology help a person perform this task more efficiently?

But:

Does a person need to perform this task at all?

For me, that’s where AI stops being another software feature.

It becomes operational.

This Isn’t Where AI Might Be Going. It’s Already Operating.

That’s ultimately why I wanted to share these results.

There’s enough speculation about what AI might eventually do.

I’m much more interested in operational evidence.

What is happening today?

What work is being executed?

What burden is being removed from staff?

And what outcomes are actually changing?

That’s the lens I believe we should use as autonomous AI moves deeper into healthcare operations.

After decades working around patient responsibility, I see the change happening.

We’re moving from software that helps people perform repetitive work to software capable of performing the work itself.

And when that happens, we shouldn’t simply automate the old process.

We should rethink the process.

That’s why I believe the conversation around AI in healthcare needs to move beyond what AI can do.

It’s time to look at what it’s already doing.