Operator's primer · 8 stages · 8 leaks
End-to-end RCM cycle: eligibility through reconciliation.
Eight stages from the moment a patient calls to the moment cash posts and reconciles. At each stage, the typical revenue leak, the AI intervention that closes it, and the measurable outcome that tells you whether the fix worked.
The premiseThe cycle is not a list of tasks. It is a compounding system.
Most revenue cycle diagrams look like a line. Patient registers, claim files, money arrives. The diagram is helpful for teaching new hires and is misleading for anyone trying to fix a leak. The cycle is a system. Each stage feeds the next. A small problem at the front of the cycle becomes a bigger problem at the back. A 2 percent leak at eligibility becomes a 4 percent denial rate at submission, becomes a 6 percent appeal volume, becomes a 90-day AR bucket that nobody can recover. By the time the leak shows up in the cash report, it has already compounded across four stages.
This primer walks the eight stages in order. For each stage, the typical leak rate, the AI intervention that closes it, and the measurable outcome. Use it as a map for diagnosis, not a script for execution. The execution is harder than the map.
A 1.5 percent leak at each of six stages compounds to 8.7 percent of net revenue. The leaks rarely live at one stage. The fix rarely lives at one stage either.
The eight stages, each with its own leak.
For each stage, the typical leak rate, the AI intervention that closes it, and the measurable outcome. Numbers are industry ranges for a mid-sized U.S. practice. Substitute your own.
Stage 01 · The front door
Eligibility and prior authorization.
The patient is on the schedule. The team verifies active coverage, captures the benefits structure, and starts any required prior authorization. The leak here is split between unverified coverage on the day of service and authorization gaps that become denials.
Typical leak2 to 4%
Outcome metricActive-coverage hit rate
AI interventionReal-time 270/271 verification at the moment of intake, same-day refresh before the visit, packet-assembly automation for prior auth.
Stage 02 · The encounter
Charge capture.
The encounter happens. The clinician documents. Charges are generated, either from the encounter form, an EHR-native template, or a coder reviewing the chart. The leak here is missed charges, late charges, and charges entered against the wrong patient or encounter.
Typical leak1 to 3%
Outcome metricCharge lag, days
AI interventionEncounter-to-charge reconciliation against the schedule, daily exception report for visits with no charges captured within 48 hours.
Stage 03 · The translation
Coding.
The documentation becomes structured codes. ICD-10, CPT, HCPCS, modifiers, and HCC mapping where applicable. The leak here is undercoding, overcoding, missed specificity, and lost HCC weight on chronic conditions.
Typical leak1 to 3%
Outcome metricRAF accuracy, coding audit pass rate
AI interventionCode suggestion with confidence scores, HCC capture prompts on chronic conditions, deterministic V28 crosswalk for risk-adjusted populations.
Stage 04 · The pre-flight
Scrubbing.
The claim is checked for completeness, payer edits, modifier integrity, and known denial patterns before submission. The leak here is the claim that goes out with a fixable error and comes back as a denial that costs ten times the effort to recover.
Typical leak2 to 5%
Outcome metricFirst-pass clean-claim rate
AI interventionPer-payer edit library, denial-pattern-trained scrubber, real-time hold for high-risk claims with a coder review queue.
Stage 05 · The handoff
Claim submission.
The claim goes to the clearinghouse, then to the payer. The leak here is small but consistent. EDI rejections, payer-side technical errors, lost batches, and silent failures that nobody catches until day 30.
Typical leak0.5 to 1.5%
Outcome metricAcknowledged-claims rate at 72 hours
AI interventionSubmission monitoring dashboard, automated batch reconciliation, exception flagging within 72 hours of submission.
Stage 06 · The pushback
Denial management.
The denials arrive. Some are reworkable. Some are appealable. Some are write-offs. The leak here is the denial that nobody works because the staffing model assumes only the high-dollar denials are worth touching.
Typical leak3 to 6%
Outcome metricAppeal overturn rate
AI interventionPer-CARC-code workflow router, appeal letter generator with payer-specific language, predicted overturn probability for prioritization.
Stage 07 · The cash arrives
Payment posting.
The cash arrives. The ERA posts. The contractual adjustments are written. The patient responsibility is flagged. The leak here is mis-posted adjustments, incorrect contractual write-offs, and patient responsibility that does not get billed.
Typical leak0.5 to 1.5%
Outcome metricPosting accuracy, percent posted in 7 days
AI interventionERA auto-posting with exception flagging, contractual variance detection against the fee schedule, patient-statement TAT monitoring.
Stage 08 · The follow-up
AR follow-up and reconciliation.
The unpaid claims age. The follow-up team works them by bucket, by payer, by aging band. The leak here is the bucket that gets touched twice and abandoned, the small-balance claim that nobody writes off cleanly, and the credit balance nobody refunds.
Typical leak2 to 4%
Outcome metricDays in A/R, percent over 90 days
AI interventionAging-band prioritization, touch-count tracking with abandonment alerts, automated credit-balance refund queue.
The compounded leak
Eight small leaks. One big number.
Each stage looks manageable in isolation. Compounded across the cycle, the eight leaks become the single largest preventable cost in revenue cycle operations.
Where AI actually helpsThe cycle compresses, not because of magic.
AI is not the right word for most of what compresses the cycle. The right word is automation with a model embedded where judgment was needed. Eligibility verification at scale is a 270/271 transaction with a payer-side response. The model layer interprets the response and flags ambiguity for human review. Prior authorization packet assembly is an EHR data pull, a payer rule lookup, and a packet generation step. The model layer chooses the right rule and assembles the packet. Coding suggestion is a chart-to-code engine with a confidence score. The model layer makes the suggestion and the coder accepts or overrides.
In every case, the AI is most useful at the boundary between data and decision. The decision still belongs to a human. The data is now ready when the human needs it, not three days later. The cycle compresses because the queue at each stage shortens, not because any single step gets magically faster.
The discipline that actually worksFix the front. Audit the back.
Most practices try to fix the back of the cycle first. Denials are loud, AR aging is visible on every dashboard, and the team running denial management is usually the most senior. The instinct is wrong. A leak at the front of the cycle creates the work the back of the cycle is busy with. Fix the eligibility leak and denial volume drops 20 to 30 percent. Fix the scrubbing leak and appeal volume drops another 15 percent. The back of the cycle gets quieter without anyone working harder.
The discipline that scales is front-first, back-audited. Fix the front of the cycle stage by stage, then run a monthly audit on the back to confirm the front-end fixes actually held. Repeat. The compounded leak does not close in a quarter. It closes over four to six quarters of steady work.
What good looks likeThe cycle in 28 days.
A mature, end-to-end RCM operation runs charge-to-cash in roughly 26 to 36 days. First-pass clean-claim rate sits at 96 percent or above. Net collection rate sits at 95 to 98 percent. Denial rate sits at 4 to 7 percent. Credentialing TAT runs at 25 to 45 days. Compounded leak across the cycle sits at 2 to 4 percent rather than 8 to 13 percent. The team running it is smaller than most practices expect, not larger.
End-to-end RCM frequently asked questions.
Quick answers to the questions practice leaders ask when they start mapping their cycle.
What are the stages of the revenue cycle?
The eight-stage view runs eligibility and authorization, charge capture, coding, scrubbing, claim submission, denial management, payment posting, and AR follow-up plus reconciliation. Some teams collapse or expand the stages, but the eight-stage map is the cleanest baseline.
Where in the revenue cycle do most leaks come from?
Eligibility and authorization at the front, coding and denial management in the middle, and AR follow-up at the back. The leaks compound. A 1.5 percent leak at each of six stages compounds to roughly 8.7 percent of net revenue.
What is the typical leak rate at eligibility?
Most practices leak 2 to 4 percent of net revenue at eligibility. The leak is split between active-coverage misses (writing off claims for patients who were actually covered) and prior authorization gaps that became denials.
How does AI compress the cycle?
AI compresses the cycle by automating eligibility checks at the moment of intake, by assembling prior authorization packets from EHR data, by running coding suggestions with confidence scores, and by predicting denial overturn likelihood. The compression is real when the AI has access to clean data and a human in the loop.
What is the right way to measure end-to-end performance?
Five composite measures: net collection rate, first-pass clean-claim rate, days in A/R, denial rate by category, and credentialing TAT. Together they capture the health of the cycle without becoming a vanity dashboard.
How long should the cycle take, charge to cash?
A mature practice should run charge-to-cash at 26 to 36 days. Teams running above 50 days have a compounded leak across multiple stages and should run a root-cause audit rather than a single-stage fix.
Should I fix the front of the cycle or the back?
Fix the front first. A leak at eligibility creates a denial at submission, which creates an appeal at denial management, which creates aging at AR follow-up. Compounded leaks have to be solved at the source. Back-end fixes are reactive.
What is the single highest-leverage intervention?
Automating eligibility and benefits verification at the moment of intake, with a same-day coverage refresh and a real-time patient financial estimate. Done well, it reduces denials at submission by 30 to 45 percent and reduces patient AR by 15 to 25 percent.
Want a stage-by-stage audit of your cycle?
A free 30-day audit under a same-day BAA. The output is a written report mapping the eight stages against your real data, with measured leak at each stage and a quarter-by-quarter remediation plan. A senior partner on the call.