The 25 percent number: where it comes from.
Family medicine practices fail to collect 25 percent of money owed every year. That is not a marketing figure. It is drawn from the University of Chicago's "A Denial A Day" study, which extrapolated the pattern to $125 billion in unpaid claims industry-wide. MGMA's 2025 Annual Regulatory Burden Report tracks the same problem from a different angle: 60 percent of denied claims never get resubmitted, 18 percent never get pursued at all, and registration or eligibility issues account for 27 percent of all denials before a coder ever sees them.
For an individual practice generating $2 million in charges annually, that 25 percent works out to roughly $500,000 in leakage per year. On a practice generating $5 million, it is $1.25 million. Six-figure revenue with no line item, no notification, and no one working it. This is why "your practice is doing fine" and "your practice is leaking" can both be true at the same time - the leakage does not show up in the P&L because it never crossed the P&L in the first place.
The leakage clusters into three categories that overlap and reinforce each other:
Leak 1: Unbilled codes (G2211, CCM, AWV, TCM).
Your providers deliver services every day that qualify for codes nobody in your practice is billing. Not because the work is not being done - because the coding and intake workflow was built before those codes existed or before the practice systematized their capture.
G2211: the complexity add-on most primary care practices are missing.
CMS made G2211 available in 2024 as the "Visit Complexity Inherent to Evaluation and Management Associated with Medical Care Services" add-on code. In plain English: an extra reimbursement for the complexity of longitudinal primary care - the value of a provider who knows the patient, manages their chronic conditions, coordinates specialist care. Applicable to a substantial share of established-patient visits at most primary care practices.
The problem: most practices are not billing it. CMS 2024 utilization data shows dramatic under-billing across primary care specifically. The reason is process, not eligibility. G2211 requires the intake form and the coder to flag it. If your practice's process was set up before 2024, it does not flag it. Every eligible visit that missed the code is billable revenue the practice earned and never collected.
Chronic Care Management (CCM): under 15 percent capture rate at most practices.
CCM (CPT codes 99490, 99439, 99487, 99489) is a Medicare-covered service for patients with two or more chronic conditions expected to last at least 12 months. It requires 20 minutes or more of clinical staff time per month coordinating care - phone calls, care plan updates, medication reconciliation. This is work most primary care practices already do informally as part of good practice.
Most eligible patients never get billed for CCM. CMS data puts the capture rate below 15 percent across primary care. The gap is process: the practice needs written consent from the patient, tracked time in the EHR, and documentation that satisfies audit. All of those are addable, but they require a workflow change most practices have not made. Adding CCM to routine capture typically produces meaningful per-provider annual revenue lift on established Medicare panels.
Annual Wellness Visits (AWV) and Transitional Care Management (TCM).
Both are Medicare services with clear billing codes and clear eligibility, both underutilized at the practice level for the same intake-workflow reason. AWV (Medicare-covered annual preventive visit) is often skipped because patients think their commercial-insurance annual physical covers it - it does not, and the two can coexist as separate billable services. TCM (transitional care management after a hospital discharge) requires contact with the patient within specific time windows and a follow-up visit within 14 days; practices that build the intake workflow to catch discharges systematically bill it consistently, practices that do not, do not.
Leak 2: Abandoned denials.
When claims come back denied, most practices give up. The MGMA 2025 numbers are stark: 60 percent of denied claims never get resubmitted. 18 percent never get pursued at all. That is not because the denials are unrecoverable. It is because working denials at scale requires more biller-hours than most practices can afford to staff.
The mechanics of an appeal look small in isolation and large in aggregate. For each denial, a biller has to: pull the 835 remittance and identify the denial reason code, retrieve the clinical documentation supporting the claim, draft a clinical justification appropriate to the payer's appeal format, submit through the payer's specific portal or fax workflow, and follow up on a 30-60 day cycle. On a practice generating hundreds of denials per month, that adds up fast. What gets triaged out are the small-dollar denials and the complex-appeal denials - the ones where the biller does the math on effort vs recovery and moves on. Over a year that adds up to five and six figures of writeable revenue that gets written off.
The other issue: some denials get denied for reasons that recur. If your Medicaid MCO consistently denies a particular code combination because of an eligibility flag your intake team is not catching, the fix is upstream (intake process) not downstream (individual appeals). Practices that never analyze denial patterns keep generating the same denials next month, and the biller keeps writing off the same ones.
Leak 3: Aged AR nobody's working.
Aged AR - accounts receivable that have been outstanding for extended periods - is the most consistent silent revenue leak in family medicine. The industry benchmark: anything past 120 days is still 65 percent collectible if someone is actively working it. But at most practices, nobody is.
The AR report gets generated. Somebody prints it or exports to Excel. It gets filed. Next month the same report gets generated. The billers are consumed with current claims - the freshest denials, the current-month posting, the patient statements going out. Aged AR is always tomorrow's problem, and tomorrow does not arrive. Every month a claim stays unworked, the recovery probability drops. By the time somebody finally looks at 240-day claims during an end-of-year cleanup, the recovery rate is under 20 percent.
The solve is not "hire another AR specialist." Most practices cannot justify the FTE for it, and the ones that can find that the workflow burns the person out within a year. The solve is continuous, prioritized aged-AR work - claims scored by recovery probability (payer, age, amount, denial history) and worked in that order, so the highest-yield ones get worked first. Whether that is done by a person with better tools or by an AI agent with an eye on human review is a resourcing question, but the underlying discipline is the same.
The Medicaid MCO multiplier.
If your practice carries a meaningful Medicaid book, the three leaks above compound.
Medicaid managed care organizations (MCOs) each have their own claim submission requirements, prior authorization rules, and appeal processes. In Texas, that means dealing with Superior, Amerigroup, Molina, and United as separate payers with separate rules for the same clinical service. MGMA data puts Medicaid denial rates at 24 percent versus 4.8 percent commercial. A Medicaid-heavy practice is leaking dramatically more per patient encounter than a commercial-heavy one - not because the care is different but because the denial and reimbursement mechanics are.
This is where automation earns its keep. A single set of appeal templates does not work across four Texas MCOs. A biller who has to remember which MCO wants what for a corrected claim is a biller who is going to make mistakes and write off claims that should have been recovered. An AI-assisted appeal process that adapts to each MCO's specific format and submission pathway removes the memory burden and levels the recovery rate across payers.
How to measure your practice's actual leakage number.
Industry benchmarks give you a range. To know your practice's specific number, you run the numbers on your own data. Here is the four-step approach.
- Pull 90 days of billing data from your PM system. Three exports: 837 claim files, 835 remittance advice, and your aged AR report as of today. Any modern PM system exports all three (eClinicalWorks, Athenahealth, Epic, NextGen, Practice Fusion, Kareo, AdvancedMD, Greenway, DrChrono, Allscripts).
- Scan chart summaries for unbilled code opportunities. For each established-patient visit in the period, check whether G2211 was billed (should be applicable to a substantial share). For each patient with 2+ chronic conditions, check whether CCM was billed that month. For each recent hospital discharge, check for TCM. Sum the missed billable opportunities.
- Analyze denial patterns in the 835 data. Group denials by denial reason code and by payer. Identify how many were never resubmitted, how many are still within the appeal window, and estimate recovery value at industry benchmarks (60-80 percent of appealable denials are recoverable in the first cycle).
- Score aged AR by recovery probability. Bucket by 30/60/90/120+ days and by payer. Apply industry recovery-rate assumptions per bucket (95 percent at 30 days, 85 percent at 60, 75 percent at 90, 65 percent at 120+). The delta between "collectible if worked" and "currently in the aged bucket" is your unrecovered opportunity.
The catch is step 2 and step 3 at scale. On a practice generating hundreds of visits per week and hundreds of denials per month, doing this manually takes a competent RCM analyst 40-60 hours per period. That is a work-week just to measure the leakage, before any recovery work starts. Which is why most practices do not measure it and rely on the industry-benchmark range (which is directionally right but hides your specific practice's biggest opportunities).
The alternative: an AI-driven audit (like the BetaQuick Practice Revenue Snapshot) that runs the same analysis in five business days on 90 days of your data and produces a one-page report ranked by recovery value.
How AI is changing revenue cycle management for family medicine.
The AI shift is not a marketing wrapper on the same category. It changes three specific things about how the RCM work gets done, and each one attacks a different leak.
1. Code capture at chart-review speed.
Traditional code capture requires a coder to read the visit note, cross-reference the documented conditions and services against billable codes, and manually flag opportunities the provider or biller missed. On a full patient panel, this is where practices choose between speed (accept what the provider coded) and precision (audit every visit). An AI agent can scan chart summaries at speed and precision at the same time - flagging unbilled G2211 opportunities per visit, unbilled CCM on eligible patients, unbilled AWV, unbilled TCM after discharge. This directly attacks Leak 1 (unbilled codes).
2. Denial appeal drafting at scale.
The bottleneck on denials is not the strategic work - a competent biller knows how to appeal. It is the drafting work: pulling the right clinical documentation, formatting it to the payer's requirements, generating the cover letter and appeal packet. An AI agent that has been trained on your practice's payer mix can ingest 835 data, identify appealable denials, and auto-generate the appeal packet in minutes instead of hours per case. The biller reviews, adjusts, and submits. This directly attacks Leak 2 (abandoned denials) because the effort per appeal drops far enough that all denials become economically workable, not just the large ones.
3. Aged AR prioritized by recovery probability.
The aged AR report by itself is a list. What makes it useful is scoring: which claims are most likely to actually collect if worked today. That scoring requires knowing the payer, the age, the amount, the denial history, and the reason. An AI agent can produce that scoring across your entire aged AR in seconds, letting one biller work the highest-value claims first instead of top-down. This directly attacks Leak 3 (aged AR) because the biller's hours get pointed at claims that will actually collect.
The combined effect: what used to require a 5-person RCM department (coder, billing specialist, AR specialist, denials specialist, manager) can run as one system with a much smaller team on top. The team is still doing judgment work, patient communication, escalations. The clerical volume happens automatically.
When to invest in AI-augmented revenue cycle management.
Three signals that make the case for a family medicine practice, any one of which is enough to justify:
- You have never audited your practice for unbilled G2211, CCM, or AWV. The first audit almost always finds a per-provider revenue lift that pays back within the first quarter.
- Your written-off denials are more than 3 percent of billed charges annually. That is a threshold above which the recovered denial revenue alone justifies the tool cost.
- Your aged AR past 120 days is more than 10 percent of total AR. A large aged bucket is the clearest sign the recovery workflow is not keeping up.
The path we recommend to any practice: start with a snapshot before you commit to a build. Run the 90-day analysis on your actual data. If the leakage number is meaningful, the recovery investment makes itself. If it is small, you have saved the money and confirmed your team is doing better than industry benchmarks.
Frequently asked.
Roughly 25 percent of money owed (University of Chicago, "A Denial A Day"). On a practice generating $2 million in charges annually, that is about $500,000 in leakage per year across three categories: unbilled complexity codes, abandoned denials, and aged AR that no one is working.
G2211 is the visit complexity add-on CMS made available in 2024 for longitudinal primary care. It is under-billed because most practices' intake and coding workflow was set up before it existed. Adding it to routine capture typically produces meaningful per-provider annual revenue lift with no additional clinical work.
Bandwidth. The appeal process for each denial takes 30-60 minutes of biller time, and on a practice generating hundreds of denials per month, working every one requires more hours than the practice can staff. Small-dollar and complex-appeal denials get triaged out and eventually written off. AI-assisted appeals change the math by cutting the drafting time far enough that all denials become workable.
Aged AR is claims and patient balances outstanding for 30, 60, 90, or 120+ days. Anything past 120 days is still 65 percent collectible if worked - but only if worked. Most practices do not have anyone assigned to aged AR, so the money quietly moves from collectible to write-off over months.
Three things: chart-scanning for unbilled code opportunities at speed, denial appeal auto-drafting so the biller's time goes to review and submission, and aged AR scored by recovery probability so effort goes to the highest-yield claims first. Together, work that required a 5-person RCM team can run as one system with a smaller human layer on top.
Chronic Care Management is a Medicare-covered service for patients with 2+ chronic conditions, requiring 20+ minutes of clinical time per month. Most practices already do this work informally but do not bill for it - CMS data shows under 15 percent capture across primary care. The gap is process (consent, time tracking, documentation), all addable without workflow disruption.
Each Medicaid MCO has its own appeal rules and submission formats. In Texas that means dealing with Superior, Amerigroup, Molina, and United as separate payers. MGMA data puts Medicaid denial rates at 24 percent vs 4.8 percent commercial - a Medicaid-heavy practice is leaking dramatically more per encounter and needs either dedicated per-MCO expertise or automated appeals that adapt to each MCO's format.
A free Practice Revenue Snapshot on your last 90 days. One-page report across all three leak categories, five business days, 15-minute walk-through. If we find nothing, you owe nothing.