TL;DR: To automate denial follow-up and payment posting for a ModMed practice, map your denial reasons and ERA remittances to rule-based workflows, let software agents work the routine items, and send only exceptions to staff. Audit your current denials and posting lag first, automate the payers and codes with the clearest fixes, connect your remittance and claim status data, set human-review rules for appeals and short-pays, and track days in A/R and posting lag to prove it works.
Why do denial follow-up and payment posting belong in the same workflow?
Both jobs start with the same document: the remittance. When a payer sends an ERA, the file tells you what was paid, what was adjusted, and why. Posting the payment and spotting the denial are two halves of reading that one file.
Most practices split them anyway. One person posts payments, another works denials, and the handoff between them is where claims go missing. A denial that sits in a posting batch for a week is a denial that ages toward a filing deadline.
The scale is large enough to justify the effort. Kodiak Solutions reported that the initial denial rate hit 11.81% of claims in 2024 across data from more than 300,000 physicians, and noted that payers eventually pay about 90% of claims. Initial denials still take real resources to overturn, which is where the labor goes.
A Premier survey of more than 500 provider organizations put the cost in dollars. It found providers spend an estimated $19.7 billion a year reviewing denials, and that about $10.6 billion of that goes to claims that should have been paid at submission.
Treating posting and denial work as one workflow gives an agent a single input (the remittance) and a clear set of outputs: post it, correct it, escalate it, or leave it alone.
Step 1: How do you audit your denials and posting turnaround?
Start with 90 days of data. You can't decide what to automate until you know what's actually happening. Pull these numbers from ModMed and your clearinghouse reports:
- Denials by payer and reason code. Group by the claim adjustment reason code on the remittance. A handful of codes usually account for most of the volume.
- Unworked denials. Count the denials with no action logged in the last 14 days.
- Posting lag. Measure the days between a remittance arriving and the payment being posted.
- Touch time. Ask your billers to estimate minutes per denial and per remittance. Rough numbers are fine.
- Write-offs. Add up what you wrote off as unrecoverable, and note how many of those were timely-filing losses.
You'll likely find the pattern is lopsided. A few payers and a few reason codes make up the bulk of the work, and several of those have a predictable fix. Those are your automation candidates.
Keep the audit spreadsheet. It becomes your baseline, and every result you report later gets compared against it.
Step 2: Which denials and payers should you automate first?
Sort denials into three tiers based on how much judgment each one needs.
Tier 1: known fix. The denial has a clear cause and a mechanical correction. Examples include missing or invalid demographic data, duplicate claim submissions, coordination-of-benefits updates, and requests for a standard piece of information you already have. An agent can correct and resubmit these, then log the action.
Tier 2: documentation needed. The fix depends on pulling a note, report, or attachment from the chart. An agent can assemble the packet and queue it for a biller to review and send.
Tier 3: clinical or contractual judgment. Medical necessity disputes, underpayments against contract, and anything that requires a physician's argument. These stay with people.
Be careful with the categories that are growing. Kodiak's data showed medical necessity denials up 5% and requests for more information up 5.4% in 2024. Both tend to land in Tier 2 or Tier 3, so don't expect automation to clear them without human review.
Start with Tier 1 for your two or three highest-volume payers. Honey Health's Denial Management and Payment Posting agents are built around this pattern: agents work the routine items, log every action, and route the exceptions to staff. Whatever tool you choose, ask to see how it handles a denial it can't classify.
Step 3: How do you connect your remittance and claim status data?
Agents need three inputs: the remittance, the claim record, and the payer's current status.
Remittance data. Most payments reach you as an electronic remittance advice, the HIPAA-standard 835. CMS describes the standard remittance advice as the file that explains payment and adjustments, using claim adjustment reason codes and remittance advice remark codes. Those codes are the raw material for sorting denials into tiers.
Claim data. The agent has to match each remittance line to the right claim in your system. Clean matching depends on consistent patient, encounter, and claim identifiers, so check this early in the pilot.
Status data. Some information never arrives in a file. The 2025 CAQH Index found 78% of medical claim payments run electronically, which means a meaningful share still arrive by paper or through payer portals. Agents that can read portals fill that gap, but portals change without notice, so ask any vendor how they detect and handle a layout change.
Confirm the access model before you sign anything. Find out what the agent can read, what it can change, and whether every action is logged where your compliance lead can review it. A signed business associate agreement is the minimum.
Step 4: What rules should send work to a human?
Automation without escalation rules produces silent mistakes. Write the rules before you turn anything on. A solid starting set:
- Short-pays above a threshold. If the payment is below your contracted rate by more than a set dollar amount or percentage, flag it for review instead of posting it as a routine adjustment.
- Unmatched payments. If a remittance line can't be matched to a claim with confidence, hold it for a person.
- Recoupments and credits. Treat negative adjustments as exceptions, since they often need a conversation with the payer.
- Appeals. Any denial that needs a letter, a clinical narrative, or a peer-to-peer call goes to a named biller.
- High-dollar claims. Set a dollar line above which a person always reviews the action.
- Anything unclassified. If the agent can't map a reason code to a rule, it stops and asks.
Assign an owner to each queue. A rule that routes to "billing" with no name attached will drift. Routing to a person with a target turnaround, such as two business days, keeps exceptions from piling up.
Review the rules monthly at first. As the agent clears more items, you'll find thresholds you can loosen and edge cases you need to tighten.
What does the daily routine look like once it's running?
The change shows up in how your billers start the day. Instead of opening ModMed and deciding where to begin, they open three short queues.
The first is exceptions from posting: short-pays above your threshold, unmatched remittance lines, and recoupments. Each item arrives with the remittance detail already pulled and a note on why the agent held it. The biller decides, acts, and closes it.
The second is denials awaiting documentation. These are Tier 2 items where the agent has gathered what it could and needs a person to review the packet before it goes to the payer. The biller's job is to check the work and send it, which takes far less time than building it from scratch.
The third is appeals and escalations: the Tier 3 cases that need judgment, a letter, or a call. This queue is smaller than the raw denial count, because the routine denials never reach it.
Meanwhile, the routine work moves without a person touching it. Payments post, known-fix denials get corrected and resubmitted, and every action lands in a log. Once a week, a lead reviews a sample of the agent's actions the same way they'd spot-check a new hire's work. If the sample looks clean, the thresholds stay. If it doesn't, the rules get tightened before the next batch.
This routine also exposes problems upstream. When the same payer keeps denying the same code, the pattern is visible in your logs, and you can take it to your front desk, your coders, or your payer rep. Denial follow-up stops being a pure cleanup job and starts feeding prevention.
Step 5: How do you measure results and handle what goes wrong?
Compare four numbers to your Step 1 baseline every week of the pilot:
- Days in A/R
- Posting lag
- Count of unworked denials older than 14 days
- Dollars recovered from resubmitted or corrected denials
Add a fifth number that's easy to forget: exceptions per hundred items. If it stays high, your rules or your data need work. If it falls steadily, your rules are catching your practice's patterns.
Plan for these failure modes:
- Payer portal changes. A redesigned portal can break a workflow overnight. Build a daily check that confirms the agent is still reading each portal.
- Unusual adjustments. Payers occasionally use codes in nonstandard ways. Keep a human review path for anything that doesn't fit.
- Source data errors. A wrong subscriber ID or missing modifier produces a bad resubmission. Fix recurring data problems at the source, in front desk or coding workflows.
- Overconfidence. Don't expand to Tier 2 until Tier 1 has run clean for a few weeks.
Run a 60 to 90 day pilot with one or two payers, report the numbers, and expand from there.
Frequently Asked Questions
Can a ModMed practice automate payment posting?
Yes. Most practices start with ERA auto-posting because the input is a structured file. An agent matches remittance lines to claims, posts payments and contractual adjustments, and flags short-pays and unmatched items for a person. Define your exception rules before turning it on.
What denials are safe to automate?
Denials with a known cause and a mechanical fix, such as demographic errors, duplicate submissions, and coordination-of-benefits updates. Medical necessity denials, contract underpayments, and appeals that need a clinical argument should stay with staff, with agents assembling documentation where possible.
How long does denial automation take to set up?
Plan on a scoped pilot of 60 to 90 days. Most of the time goes to access, data matching, and writing exception rules. Starting with one or two payers and the Tier 1 denial types keeps the pilot manageable.
Will automation catch timely filing deadlines?
It can help. An agent that works denials daily and flags aging items shortens the time a fixable denial sits idle. You still need a rule that escalates anything approaching a payer's filing limit to a named person.
What should I track to prove it's working?
Track days in A/R, posting lag, unworked denials older than 14 days, dollars recovered, and exceptions per hundred items. Compare each to the baseline you collected before the pilot, and report them weekly.
