Six Figures, Two Weeks

The Field Report

In April 2026, HRN Group completed the first two weeks of a live engagement at a Kansas rural hospital. The hospital is independent and community-anchored, operating against the four KanCare managed care contracts plus Kansas Medicaid fee-for-service. Roughly half of its payer mix is Kansas Medicaid in some form. Its revenue cycle team is small. Its operating margin is the kind that makes uncompensated care a survival-grade variable, not a back-office annoyance.

In fourteen days of operations, we recovered six figures of underpaid Medicaid revenue back into the hospital’s account. Not “identified.” Not “projected.” Recovered, into deposits, traceable on the hospital’s books.

This white paper is not a marketing case study. It is a methodology document. We are publishing it because every rural hospital CFO operating against the same Kansas Medicaid managed care contracts is sitting on the same recoverable revenue, in patterns that repeat with high consistency across hospitals on the same payer footprint. We want CFOs, hospital association leaders, and Medicaid MCO operations leaders to understand what is actually possible, what we actually did, and how this differs structurally from horizontal AI RCM products.

The hospital’s name, specific recovery dollar amounts, and individual staff identities are anonymized per the active Business Associate Agreement covering the engagement. Every methodology element, timeline, and pattern described in this paper is exactly as it happened.

This paper makes three arguments built on that field engagement:

• Rural hospital revenue recovery is, in practice, a visibility problem first and a recovery problem second. Most of what we recovered was sitting in claims whose adjudication detail had never been ingested back into the hospital’s accounting system in a form the staff could see. The denials were not unrecovered because the staff failed to fight them. They were unrecovered because the staff couldn’t see them.
• A three-layer methodology — data ground-truthing, pattern recognition at scale, recovery operations — compresses the standard 90-day vendor deployment cycle to a 14-day operator cycle. The compression comes from skipping the platform-rollout layer entirely and embedding into the hospital’s existing workflow, augmenting its billing team rather than replacing it.
• The patterns generalize. The hospital is one site. The KanCare contracts apply across every hospital in Kansas operating against KS Medicaid managed care. The MCO behaviors, code drifts, and adjudication anomalies we surfaced are payer-side behaviors, not hospital-side errors. They repeat at scale.

What follows is the methodology in detail, the operational evidence behind these arguments, and a clear-eyed comparison to the alternatives a rural hospital CFO is likely evaluating.

The Visibility Problem

The standard framing of rural hospital revenue recovery — the framing every horizontal AI RCM vendor uses, and the framing most CFOs internalize — is that there is a denial recovery problem to be solved. Claims get denied. Staff don’t have time to fight every denial. AI tools can prioritize, automate appeal generation, and scale denial fighting. Result: more denials worked, more revenue recovered.

Our experience at this engagement demonstrates that the framing is wrong, or at least incomplete. The hospital did not have a denial recovery problem. They had a denial visibility problem. The two are not the same problem, and treating them as the same produces tools that solve the wrong layer.

Here is what we actually found in the first three days.

The hospital’s billers were working denials they could see — the ones that surfaced in their patient accounting system, queued for follow-up. They were closing the highest-value cases professionally and triaging the rest. When we looked at what was upstream of their work queue, we found a meaningful population of claims that had been submitted, adjudicated by the payer, and short-paid — but whose adjudication detail had never been ingested back into the patient accounting system in a form the staff could see. The denials existed in the 835 remittance files. They existed in the clearinghouse logs. They were not in the system the billers were working.

The denials were invisible to the team that would have fought them.

This visibility gap is structural, not idiosyncratic. The EHR-to-clearinghouse-to-accounting-system data path in most rural hospitals is held together with batch processes, vendor handoffs, and historical configuration choices that nobody has had time to revisit. Some 835 fields don’t get parsed. Some claim adjustments get mapped to ambiguous codes that look “closed” rather than “underpaid.” Some MCO portal denials never make it into the EDI 835 stream at all. Each individual gap is small. Aggregated across a year of claims, the gap becomes financially meaningful.

This is not a criticism of the hospital. The data path is structurally broken across the rural hospital landscape because no individual hospital has the IT bandwidth to fix it. It is a vendor problem, an integration problem, and a configuration problem — none of which a small billing team can solve while also working denials by hand.

The diagnostic implication is significant. If the recovery problem is downstream of a visibility problem, then no amount of denial-fighting AI applied at the recovery layer will reach the unrecovered revenue. The denials the AI is working are the denials the hospital’s staff was already working — possibly faster, possibly with better appeal letters, but the same set. The denials that matter, the ones sitting invisible in the data path between adjudication and accounting, never enter the AI’s input queue.

This is why the standard horizontal AI RCM deployment timeline is 60 to 90 days. Most of that time is consumed by platform integration, not by recovering money. And much of that integration leaves the visibility gap intact — the platform reads from the same accounting system the staff was already reading from.

The fix for the visibility problem is not more denial-fighting capacity. It is data ground-truthing, performed by operators who know what to look for. Once the visibility layer is restored, the recovery problem becomes manageable for a small competent billing team augmented by analytical tooling that scales pattern recognition across the now-complete claim corpus.

The Three-Layer Methodology

The HRN engagement model has three distinct layers, deployed in sequence and in parallel where dependencies allow. Each layer answers a different operational question. The compression of the 90-day cycle into a 14-day cycle comes from doing all three with operator-level focus rather than platform-level breadth.

Data ground-truthing

Goal: reconcile the four data planes that should agree but rarely do.

Most rural hospitals have at least four distinct data representations of the same claim:

• The claim as submitted (837 file, EHR claim record, clearinghouse log)
• The claim as adjudicated (835 remittance file from the payer)
• The claim as ingested (patient accounting system record)
• The claim as worked (denial queue, follow-up notes, appeal documentation)

In a healthy revenue cycle, these four planes agree. In every rural hospital we have engaged, they do not. Discrepancies fall into predictable categories: 835 lines that didn’t ingest cleanly because of code mapping gaps; adjustment reasons that ingested but mapped to ambiguous status flags; portal-only denials that bypass the EDI stream entirely; manual write-offs that mask underpayments rather than expose them; resubmissions that recreate the same problem in a new claim record.

The ground-truthing layer reconciles these four planes line by line for the in-scope claim corpus. We do this work alongside the hospital’s billing team — not over their heads — using their data, their accounting system access, and their domain knowledge of payer-specific quirks. The output is a complete, deduplicated, accurately-coded claim corpus on which subsequent analysis can run.

This is the layer most horizontal AI RCM tools skip. They trust the data they’re handed and run pattern recognition on top. The recommendations are systematically wrong because the underlying corpus is incomplete. We start by ensuring the corpus is right.

Pattern recognition at scale

Goal: surface payer-side patterns no manual review process can catch in production volume.

Once the data is clean, we run agentic AI against the full claim-and-remittance corpus to identify patterns clustered by payer, by code, by submission date, by adjudication date, and by adjustment reason. The agents are not being asked to “find denials worth fighting” — they are being asked to find statistical anomalies in payer behavior that suggest systematic underpayment.

In our Kansas engagement, three categories of pattern emerged within the first week:

CARC/RARC code drift. Specific MCOs were issuing denial reason codes that did not match the published payer guidance for the underlying clinical scenarios. The code shifts clustered around quarterly fiscal-year-end submission windows — strong evidence of payer-side policy changes that were not communicated through the formal provider notification channel. The hospital’s billers had been responding to each denial individually as a coding question. The pattern across hundreds of denials was the actual signal: these were not coding errors, they were payer-side policy drift.

Adjudication anomalies on identical codes. The same procedure code, billed in identical fashion, paid correctly on Tuesday and short-paid on Thursday at the same MCO. There was no medical, billing, or contractual reason for the variance. The variance was concentrated at two of the four KanCare MCOs and was almost certainly the result of payer-side AI logic the hospital had no visibility into. We documented the pattern at sufficient detail that it could be filed as a reconsideration with explicit pattern documentation rather than as individual case-by-case appeals.

Kansas Medicaid–specific underpayments on outpatient codes. A population of underpayments that appeared only on Kansas Medicaid claims — not on Medicare, not on Medicare Advantage, not on commercial. The KS Medicaid pricing logic was producing payments below the contracted fee schedule on specific outpatient services, with no acknowledgment of the underpayment from the payer side. The pattern is appealable but requires state-specific documentation that takes more elapsed time than a payer-side reconsideration.

These categories were not what the hospital expected to find. The hospital expected the work to surface “the usual” — coding errors, missing documentation, late filings. What we surfaced was payer-side behavior the hospital had no internal mechanism to detect and no individual-claim path to correct.

Recovery operations

Goal: convert visible patterns into recovered dollars in the hospital’s account.

Once patterns are identified, recovery is no longer an analytical problem. It is a documentation, submission, and follow-up problem. We work alongside the hospital’s billing team to file corrected claims for clean coding errors, file reconsiderations for adjudication anomalies with explicit pattern documentation, file first-level appeals for documented underpayments where the payer position is that the original payment was correct, track each cycle through to either recovery or terminal denial, and maintain the audit trail across the recovery work in a form that survives downstream review.

The hospital’s billers stay in their seats. Their workflow doesn’t change. They send the same claim formats, work in the same accounting system, and use the same payer relationships they have always used. What changes is the analytical air cover behind each individual fight: instead of fighting one denial at a time, they are fighting one denial in the context of a documented payer-side pattern that affects dozens or hundreds of similar claims.

The leverage is significant. A reconsideration filed with pattern documentation gets a meaningfully different reviewer response than a reconsideration filed as a one-off. The payer side knows the difference between a hospital that has noticed a pattern and a hospital that is fighting individual cases.

The Sprint Cadence Model

We organize recovery work in two-to-four-day operational sprints, each defined by a scope (which subset of claims), a pattern set (what we’re attacking), and a recovery target (what we expect to land). Sprints run in parallel where the dependencies allow and in series where they don’t.

In the field engagement, four sprints ran across the fourteen-day window:

High-confidence underpayments, KS Medicaid FFS

Claims paid below the published KS Medicaid fee schedule. These are the easiest cases — patterns are unambiguous, documentation is straightforward, recovery is a matter of corrected claim filing. Five-figure recovery.

CARC/RARC-coded denials, KanCare MCOs

Claims denied with codes the published guidance did not predict. Recoveries through corrected resubmissions and reconsiderations with pattern documentation. High five-figure recovery.

Adjudication anomalies, two named MCOs

The Tuesday-Thursday pattern. Reconsiderations filed with explicit cross-claim pattern documentation. Sprint 3 was harder — MCO response cycles longer, documentation work heavier, some cases required first-level appeal escalation. Mid five-figure recovery.

Outpatient-code underpayments, KS Medicaid

The hardest of the four. Underpayments where the payer position was that the original payment was correct, requiring formal appeal with state-specific documentation. Mid four-figure recovery so far — appeal window is longer, more of this work is still in flight.

Total recovered through fourteen days of operations: six figures, into the hospital’s account.

Material in flight beyond the two-week window: another full sprint’s worth of cases at appeal stage, plus a second-pass review of the data ground-truthing work that surfaced additional patterns we did not have time to chase in the first two weeks.

The sprint cadence is the operational backbone that makes the 14-day timeline viable. By scoping work into discrete sprints with clear targets, the engagement avoids the trap most enterprise RCM rollouts fall into: a multi-month integration phase before any revenue is recovered. The first sprint produces recovered revenue in days, which funds the engagement, validates the approach, and gives the hospital’s CFO a defensible internal narrative for continuing the work.

Why Generic AI RCM Cannot Replicate This

This section is unavoidably blunt. We are competing against a category of vendor that has raised hundreds of millions of dollars on the promise that AI alone solves rural hospital revenue cycle. The Peterson Health Technology Institute report we covered in our companion paper, When AI Makes the Problem Worse, documented why that promise has not held up in field deployments. Our experience at this engagement confirms PHTI’s findings from the operator side.

Three structural reasons horizontal AI RCM products cannot replicate the methodology described above:

They skip the data ground-truthing layer

The platform-deployment model presumes the data the platform reads is the data that exists. The visibility gap that contains most of the unrecovered revenue is upstream of the platform’s input. An AI deployed inside the patient accounting system cannot see denials that never reached the patient accounting system. The 60-to-90-day deployment timeline is consumed by platform integration; the visibility gap remains.

They treat staff as a constraint rather than a resource

Enterprise AI RCM products are designed to scale by reducing dependence on staff. The implicit goal is automation — the human is the thing being replaced. In rural hospitals, this is the wrong model. The small billing team is not the bottleneck; it is the institutional knowledge of payer behaviors, internal documentation pathways, and clinical reality. Augmenting the billing team is a higher-leverage move than replacing it. Operator-led engagements augment; platform-led engagements replace. The augmentation model produces compounding returns. The replacement model produces churn.

They optimize for breadth, not depth

A horizontal AI RCM product needs to work across every hospital, every payer, every state, every contract structure. That generality forces the product into average-case behavior — surfacing the patterns that are common across the entire installed base. The patterns we surfaced in the field engagement were Kansas-Medicaid-specific, MCO-specific, and contract-specific. A horizontal product cannot afford to develop the depth required to identify those patterns; an operator focused on a specific market can.

Generalizing Across the Rural Hospital Population

If you operate a Kansas rural hospital running against KanCare and KS Medicaid fee-for-service, three things are probably true at your hospital right now, with high confidence based on the patterns we surfaced:

You have recoverable Medicaid revenue sitting in claims you cannot see. Not because your billing team is bad — because the visibility gap is structural across the EHR-to-clearinghouse-to-accounting-system data path that almost every rural hospital runs.

The patterns we identified repeat across hospitals on the same contracts. CARC/RARC drift at named MCOs is payer behavior, not a hospital-specific quirk. Adjudication anomalies at named MCOs are payer-side AI logic, not a coding error in your shop. KS Medicaid outpatient underpayments are state-side pricing logic, not a billing issue in your accounting system. If they are happening at one rural hospital on these contracts, they are almost certainly happening at yours.

The recovery is a fourteen-day game, not a 90-day game — when the methodology described in this paper is followed. We do not claim the timeline is universal. We claim it is achievable when the engagement model fits.

What this means for a CFO evaluating whether to invest in revenue recovery work:

The dollar-figure question — “how much recoverable revenue do I have?” — is the wrong first question. The right first question is “is my visibility layer complete?” If the answer is no, then the dollar figure is necessarily larger than what your current denial queue suggests, and the path to recovering it begins with ground-truthing rather than with denial-fighting.

The dollar-figure ranges to plan against, based on the field engagement and structurally similar work elsewhere in the rural hospital population:

• Hospitals with 30%+ KS Medicaid in payer mix can plan to recover meaningful five-to-six-figure dollar amounts in the first 14 to 30 days of properly-conducted operations. The work after that continues to surface recoverable patterns at a slower but consistent cadence.
• Hospitals with high Medicare Advantage exposure see analogous patterns on the MA side, with structurally similar but operationally different recovery paths.
• Hospitals with diversified payer mix and lower Medicaid exposure see smaller absolute recoveries but proportionally similar visibility gaps.

The numerical specifics are necessarily a function of the individual hospital’s payer mix, claim volume, and contract structure. The structural argument is universal: if your visibility layer is incomplete, you have unrecovered revenue. The methodology to surface and recover it is described in this paper.

The MCO Side of the Same Engagement

We are publishing this paper knowing the four KanCare MCOs whose adjudication patterns we surfaced will read it. We want them to.

Two things we want named MCO operations leaders to know.

Most of what we recovered was money your plan would have paid correctly if the upstream data and documentation had been clean. The portion that came through reconsideration of misadjudicated claims is real, and we will fight for it on every site. The larger portion came through corrected resubmissions and through ingestion-gap visibility restoration. Cleaner data flowing both ways means less rebill churn for both sides. The PHTI report described that churn as one of the largest manual labor cost categories inside Medicaid managed care. We agree. The way to reduce it is upstream data quality, not downstream automation.

When CMS-0057-F goes live on January 1, 2027, the providers who have already done this kind of upstream work will be the partners you want. A FHIR-PAS-fluent provider on the other side of the prior auth and claims interface is a provider whose data your reviewers can trust the first time. That is the dynamic we are building HRN for. We are happy to talk to MCO operations leaders about what that partnership looks like before the regulatory deadline forces the conversation.

Engaging HRN

We do not engage every rural hospital that asks. We engage where the methodology fits, where the hospital’s leadership is committed to working alongside us, and where the payer mix supports sprint-cadence recovery operations.

The engagement starts with a 20-minute conversation. We tell you whether we are the right fit, and if we are not, we tell you who is. There is no charge for the conversation, no pitch deck, and no commitment to proceed. The most valuable outcome from many of these conversations is a realistic assessment of where the hospital’s revenue cycle actually stands relative to other peer hospitals on the same contracts.

If the fit is right, we propose a defined scope-and-target engagement. The first sprint produces recovered revenue inside the first two weeks. The engagement scales from there based on what we find.

Our pricing is recovery-share for hospital engagements — we are paid as a percentage of recovered dollars, with a defined share of revenue recovered through the engagement. This aligns our economics with the hospital’s economics. If we don’t recover, we are not paid.

The bottom line

Rural hospital Medicaid revenue is recoverable on a fourteen-day timeline when the methodology fits the hospital. The bottleneck is not technology. The bottleneck is not staff capacity. The bottleneck is the visibility gap between adjudication and accounting that no individual hospital has the IT bandwidth to fix on its own. HRN exists to fix that gap and convert the visibility into recovered dollars. The field engagement described in this paper is the proof. The methodology is the playbook. The next site is a conversation away.

To start that conversation, reach out here. We will respond within one business day.