How Nomad Data's Doc Chat Helped RKL Cut Claims Review Time by 70%

RKL is a leading advisory firm with more than 700 professionals working across 30 states. Its services span tax, accounting, financial management, workforce strategies, private wealth, technology, and more, helping organizations and individuals prepare for what comes next.

That forward-looking approach extends to RKL's Senior Living Services Consulting Group, where medical billing teams manage complex reconciliation work across claims and supporting therapy documentation. Before Nomad Data's Doc Chat, experienced reviewers compared claims against supporting records one page and one data point at a time. The work was necessary, but repetitive, time-intensive, and difficult to scale as file volume increased.

Today, the RKL Senior Living Services Consulting Group uses Nomad Data's Doc Chat to complete the first comparison, identify possible discrepancies, and direct reviewers to the items that deserve attention. RKL reports that the redesigned workflow reduced its Part B claims review time by 70%.

The result did not come from replacing skilled billing professionals. It came from redesigning the division of labor. AI handles repetitive matching and rule-based checks. People remain responsible for validation, interpretation, and deciding what happens next.

RKL’s story is a practical model for applying AI to a high-volume, rules-driven workflow: start with a narrow use case, keep experts in the loop, train the system with real examples, and measure the capacity returned to the team.

Why medical bill reconciliation is harder than it looks

Medical bill reconciliation can sound like a simple matching exercise. In practice, the information being compared often comes from systems and documents that were never designed to align. A UB-04 claim may need to be evaluated against a therapy log, plan of care, authorization record, or clinical documentation. Each source can use different naming conventions, date logic, code structures, and levels of detail.

That makes the work interpretive. Reviewers are not only checking whether two fields are identical. They are determining whether the records describe the same patient, service, provider, and episode of care, and whether a difference represents a true billing problem or a normal system variation.

‍“Medical bill reconciliation looks simple on the surface, but in practice it involves matching information across sources that were never designed to line up perfectly.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

A complete review can include occurrence codes, diagnosis codes, discipline codes, start-of-care dates, certification dates, physician names, National Provider Identifier numbers, units, service dates, and billing amounts. Every field can carry downstream consequences. A seemingly minor inconsistency may lead to a rejection, payment delay, denial, or a question that sends the team back through the file.

The old workflow: inspect every claim, field, and page

Before the new workflow, RKL’s management team manually compared every claim to the corresponding therapy log. Reviewers moved through the files in detail, confirming that services, dates, units, provider information, and other claim elements agreed with the supporting therapy documentation.

The process could take hours per file. Because the same careful attention was applied to routine matches and true exceptions, highly experienced staff spent substantial time proving that correct information was correct. High volume amplified the problem. The more repetitive the review became, the more likely it was that a small but important difference could escape notice.

‍“In high-volume billing environments, the biggest risk is that reviewers are working through so much repetitive detail that small inconsistencies are easier to miss.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

This is a familiar pattern across document-heavy operational work. Thoroughness is essential, but reviewing every item with the same intensity does not always produce the best use of expert time. The opportunity was to separate routine agreement from the exceptions that required judgment.

Why traditional OCR was not enough

RKL had previously explored traditional optical character recognition. The technology could extract text, but extraction alone did not solve the reconciliation problem. Source documents varied too much in format and organization, and the workflow required contextual comparisons rather than a simple transfer of text from one place to another.

The distinction matters. OCR answers the question, ‘What text appears on this page?’ Reconciliation asks a harder set of questions: ‘Does this claim agree with the supporting record? If not, which difference is meaningful? Which rule applies? What should a reviewer inspect next?’

RKL saw a stronger fit for AI because the workflow combined high volume, consistent review rules, recurring exception types, and a clear role for human oversight. The goal was not open-ended automation. It was a disciplined first pass that could apply the same checks across every file and present reviewers with a focused set of possible issues.

The new workflow: AI first pass, human review where it matters

RKL now uploads claims and supporting therapy logs into Doc Chat. The system completes the initial comparison and flags potential errors, inconsistencies, and mismatches. The reviewer then validates the output and investigates the exceptions rather than reconstructing the entire comparison manually.

‍“The reviewer’s role has shifted from conducting a full manual review of every claim to validating Doc Chat’s results and addressing only items flagged for potential errors or mismatches.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

The workflow also expanded beyond direct comparisons between claims and therapy logs. RKL added checks for information that may not appear in the therapy documentation but still matters for claim readiness, including:

• Whether the type of bill is appropriate

• Whether managed care claims include required authorization codes

• Whether insurance identification numbers are present and consistent

• Whether service dates, units, codes, and provider information align across the file

• Whether a discrepancy should be treated as a true exception or a normal variation

This exception-first design changes the economics of review. Instead of asking managers to spend equal time on every record, it concentrates their attention on the small subset of items most likely to affect accuracy, completeness, or payment.

‍“The most significant time savings comes from allowing management to focus on exceptions rather than reviewing every page and data point individually.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

The outcome: a reported 70% reduction in Part B claims review time

‍“We can confidently say...we have reduced our Part B claims review process by 70%.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

For RKL, the significance of the 70% figure is not only faster completion. It represents capacity returned to the department. Managers can spend more time on higher-value work, while senior billers can take on a larger role in the review process. The organization gains room to grow the workload without requiring expert attention to scale linearly with every additional page.

What one mismatch can cost

The value of reconciliation becomes clearest when a small discrepancy creates a large amount of downstream work. RKL pointed to a claim listing the wrong physician as one example. The claim may not match the therapy log, plan of care, or clinical record. If the inconsistency is not caught before submission, it can contribute to a denial or delay.

Resolving that single issue can require the billing team to identify the denial reason, reopen the supporting documentation, confirm the correct physician, update the claim, resubmit it, and follow up with the payer. Payment is delayed, and multiple employees spend time correcting an avoidable problem.

That is why the best use of AI in reconciliation is not simply faster document search. It is earlier and more consistent exception detection. By surfacing a mismatch before submission, the workflow can help prevent rework and protect the team’s time.  

Human judgment stays in the loop

RKL’s approach is intentionally human-led. AI performs repetitive comparisons and flags items for review, but experienced billing professionals determine whether the difference matters, what additional documentation is needed, and how the issue should be resolved.

‍“AI helps to take the first pass of reconciliation, and our team takes on the role of reviewer. It does not replace our judgment, but it helps focus that judgment where it matters most.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

This division of labor is particularly important in healthcare billing. Payer requirements differ. File formats vary. Coding rules change. Protected health information must be handled in an appropriate, HIPAA-compliant environment. A useful solution must therefore combine consistency with traceability, security, and a clear path for human verification.

‍“Using AI with judgment is the future.”  - Yenma Herb, Systems Innovation Sr. Manager, RKL

That principle is broader than medical billing. In any regulated, document-heavy workflow, the safest and most useful role for AI is often to organize the evidence, perform repeatable checks, and show people where to look. Final responsibility stays with the professional who understands the operational and regulatory context.

The real ROI is capacity

Time savings are easy to understand, but RKL defines the return more strategically. The previous review process blocked substantial portions of management time. By narrowing the work to exceptions, the new workflow gives those managers room to focus on client needs, operational improvement, staff development, and other high-value responsibilities.

‍“The biggest ROI is creating capacity for our medical billing managers to be able to focus on higher-value work.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

Capacity is a useful way to evaluate automation because it connects efficiency to business outcomes. A faster process can reduce turnaround time, but the larger benefit may be what the team does with the hours it gets back. RKL can distribute work more effectively, involve senior billers in review, and preserve management attention for decisions that require deeper expertise.

This is also why the strongest AI use cases tend to have both volume and repetition. If the same comparison happens frequently, the rules are reasonably stable, and the source information is structured enough to evaluate, a consistent first pass can create meaningful leverage. If every case is entirely unique or the underlying data is unreliable, the value will be harder to realize.

RKL’s advice: start narrow and expect iteration

RKL recommends beginning with a focused pilot rather than trying to automate an entire billing operation at once. One payer, one claim type, or one recurring comparison can provide enough scope to test the workflow while keeping the feedback loop manageable.

‍“Start with a pilot. Identify one payer to begin with. Collect examples, understand the trends and patterns, and find a vendor you trust who is willing to learn the work alongside you.”  - Lacy Albright, Practice Leader, Medical Billing, RKL

Real-world examples are essential. Teams need to identify the discrepancies that recur, distinguish true exceptions from harmless variations, and give the system feedback when its output is incomplete or incorrect. The first version should be treated as the beginning of operational learning, not the finish line.

‍“Test, evaluate, test, evaluate and test and evaluate again. AI is not going to get it perfect from the start; it’s going to be an iterative process.”  - Yenma Herb, Systems Innovation Sr. Manager, RKL

That candid expectation is one of the most important lessons in the story. AI adoption succeeds when a team owns the workflow, reviews the outputs, and improves the checks over time. RKL’s 70% result followed changes to both the technology and the way people used it.

How to identify a strong reconciliation use case

For medical billing leaders evaluating AI, RKL’s experience points to three practical readiness factors:

1. Volume: The review occurs often enough that reducing repetitive effort will return meaningful capacity.

2. Repeatability: The team applies recognizable rules and sees recurring exception types across files.

3. Data quality: The necessary information exists in the documents and can be interpreted consistently enough to support comparison.

4. Governance: Teams should know where files are processed, how information is protected, who can access the system, and how a reviewer can verify each answer. These controls are especially important when documents contain protected health information.  

Not every manual process should be automated. The best starting point is a workflow where AI can augment a skilled reviewer without obscuring the evidence or removing professional accountability.

A repeatable model for document-heavy operations

RKL’s experience offers a clear operating model for AI medical bill reconciliation. Centralize the relevant claims and supporting records. Define the checks that matter. Let AI complete the repetitive first pass. Route possible discrepancies to people. Capture reviewer feedback. Improve the workflow as new patterns appear.

The model works because it aligns technology with the actual constraint. RKL did not need a system that could make every billing decision independently. It needed a system that could reliably narrow thousands of possible comparisons into a manageable set of exceptions.

That shift, from page-by-page to exception-first, is what turned AI from a promising tool into measurable operating capacity.

Make Complex Claims Review Faster

See how Nomad Data’s Doc Chat helps insurance teams compare documents, identify discrepancies, and focus expert attention where it matters most. Explore Doc Chat for Insurance.