Autonomous Coding Drove a 5.1% Revenue Increase at Mercyhealth — Here's What Actually Happened

Mercyhealth was processing 130,000 charts a month with a coding team that couldn't keep up. They implemented autonomous coding — and got results worth paying attention to.

Mercyhealth, a 200-location health system spanning Wisconsin and Illinois, had a problem that most growing health systems know well: chart volume was outpacing coding capacity.

At more than 130,000 charts per month, the coding team couldn't keep up. Encounters were moving forward without full review. Revenue integrity was inconsistent. The team was stuck in reactive mode — bouncing between denials, audits, provider questions, and new service lines with no clear path to getting ahead of the work.

The answer wasn't more coders. It was autonomous coding.

The Setup

Mercyhealth partnered with Arintra to embed autonomous coding directly into their Epic EHR and existing revenue cycle workflows. No new system to learn. No parallel process to manage. Same workflows — with AI handling the high-volume, routine coding that was consuming the most time.

Kelly Pierson, Director of Coding and CDI at Mercyhealth, described the challenge clearly:

"This often meant shifting focus between urgent needs, making it difficult to stay ahead of the work or operate as proactively as we would have liked. It became clear the demand for the team was continuing to grow and it wasn't something we could sustainably manage with traditional approaches alone."

They started small — a pilot with a select group of family and internal medicine providers. Validated the model. Built confidence. Then expanded across specialties. Smart rollout. Most failed AI implementations skip this step entirely.

Before go-live, they also standardized. Templates aligned. Coding policies made consistent across encounter types. Arintra worked with the team to establish that baseline before automation touched any volume. Garbage in, garbage out still applies.

The Results

The 5.1% revenue increase came from more accurate, complete capture of the care being delivered — not from billing more aggressively, but from billing correctly the first time.

What "Higher-Value Work" Actually Means

This is where most AI coding coverage gets lazy. The narrative defaults to "coders will lose their jobs" or "AI will replace humans." Neither happened at Mercyhealth.

What happened instead: the coding team stopped chasing volume and started driving strategy.

"Our coding team is spending more time on complex cases and higher value work where their expertise is most needed. They are also able to focus more on areas like denial trends, revenue integrity and provider education." — Kelly Pierson, Director of Coding and CDI

That's a meaningful shift. Coders with deep clinical knowledge are most valuable when they're analyzing denial patterns, educating providers on documentation gaps, and identifying revenue integrity issues — not manually reviewing routine E/M visits that a well-trained model can handle consistently.

The 50% Pre-A/R Reduction Is the Underrated Metric

The 5.1% revenue increase will get the headlines. But the 50% reduction in pre-A/R days is arguably more operationally significant.

Days in A/R measures how long it takes to convert a service into cash. Cutting that by half — even for a subset of encounters — has compounding effects on cash flow, working capital, and the ability to operate without a buffer reserve.

Cleaner claims on first submission means fewer touchpoints, fewer delays, and less rework downstream. When denials do occur, the team now has clearer documentation-tied rationale to support appeals — no routing cases back through multiple touchpoints before a response goes out.

Pierson noted they've also been able to address smaller-dollar denials that previously weren't getting prioritized. That's incremental revenue that compounds over time.

What Mercyhealth Did Right

This implementation worked because of four things most orgs skip:

1. Standardization before automation. Aligned templates and consistent coding policies were established before go-live. The model had a clean baseline to work from.
2. Embedded workflow, not parallel process. The system lived inside Epic. No dual-entry, no switching costs, no adoption friction.
3. Human override stays intact. Complex or uncertain cases still route to a coder. Autonomous doesn't mean unilateral.
4. Measured rollout. Pilot → validate → expand. Not big-bang implementation with a hard cutover date.

Bottom Line

If your organization is processing six-figure monthly chart volumes with a coding team that's stretched thin, the question isn't whether autonomous coding is ready for prime time. Mercyhealth's results answer that.

The question is whether your team has the workflow discipline, EHR integration approach, and rollout patience to make it work.

The 5.1% revenue lift is real. But the structural efficiency — cleaner claims, faster cash, coders focused on complexity — is what makes the investment durable.