Home Use Cases Work Insights About Contact
Back to Our Work

How a Healthcare Provider Built a Claims Pipeline That Codes, Checks and Sends with a Coder in the Loop

Healthcare3
Open the walkthrough

Six-Step AI Workflow with Human Oversight

Each step shows the workspace, what the AI completed, and what still needs human review.

Steps 2, 4, 5 and 6 need a person. Step 4 is waiting for approval right now. Click it to open the full walkthrough.

Estimates are directional and based on stated assumptions. All names, organizations, and identifying details have been anonymized in accordance with our confidentiality agreements.

The Transformation

How a Healthcare Provider Built a Claims Pipeline That Codes, Checks and Sends with a Coder in the Loop

Before
A coder used to read every chart note against what was charged.
A biller used to test every claim against the payer's edits by hand.
After
every closed encounter read, and only what the note supports carried forward
codes suggested with their evidence, waiting for a coder to sign
every claim tested against the payer's own rules before it is sent
the short queue of claims only a person can release
every denial sorted, and the appeals worth writing already drafted
Executive Summary

Reads what the clinician documented, codes it, checks it against the payer's own rules before it is sent, and stops at the claims a person has to decide. Written for a regional healthcare system where the revenue cycle was losing to rework rather than to volume: 200,000 claims a month across 12 facilities, 72 out of every 100 paid first time, and the other 28 coming back to be done again. Six stations, run as a loop: what station 6 learns from a quarter of denials becomes the rule station 3 checks against.

What Was Broken

A coder used to read every chart note against what was charged
The real cost
A biller used to test every claim against the payer's edits by hand.

What We Built

Six stations and 13 subagents. Each subagent carries its own tasks and its own refusal.

A4
Encounter Watch
Poll the practice management system for every encounter closed overnight
A3
Chart Reading
Extract the services, diagnoses and modifiers the note documents
A2
Code Suggesting
Summarize each encounter into the codes its documentation supports
A3
Modifier Checking
Check each modifier against the payer's own published policy
A3
Rule Testing
Test each claim against the edits this payer publishes
A3
Risk Scoring
Score each claim on how often this payer denied one like it
A2
Risk Reading
Read each held claim back against the rule that held it
A2
Hold Briefing
Summarize each held claim in one line the decision can be made from
A4
Reason Sorting
Sort every returned claim by the reason code the payer sent
A1
Appeal Drafting
Draft the appeal this reason code has succeeded on before
A4
Pattern Counting
Count denials by payer, reason and service over the rolling quarter
A3
Payer Watching
Search each payer's published bulletins for requirements that changed
A1
Rule Proposal
Draft the validation rule this quarter's denials support

How It Runs

1

Charge Reading

Nobody is asked anything here

Subagents read every closed encounter, pull out the services, diagnoses and modifiers the note supports, and hold anything the note does not carry. Nothing waits on a person here. A charge reaches a coder only when station 2 cannot settle it.

2

Code Assigning

A person answers here

Subagents suggest the codes the note carries and show the line each one came from, so a certified coder reviews a suggestion instead of starting from an empty field. A coder signs every code. Nothing is submitted on a suggestion alone.

3

Claim Checking

Nobody is asked anything here

Subagents test every coded claim against the payer's own edits, check that nothing required is missing, and score how likely this payer is to deny one like it. A person sees only the claims the pipeline expects to lose. Everything it expects to be paid goes out.

4

Claim Releasing

A person answers here

Subagents read every held claim back against the rule that held it, draft the correction the rule supports, and stop. This is the station where a human works. Everything else exists to make this decision small and well-lit.

5

Denial Working

A person answers here

Subagents sort every returned claim by its reason code, separate what can be corrected from what cannot, and draft the appeal that has worked on this reason before. A person decides what to appeal. The draft is ready either way.

6

Payer Learning

A person answers here

Subagents count every denial by payer and reason, watch what each payer has published since, and propose the validation rule that would have caught it before it was sent. A person accepts or rejects each proposed rule. Nothing about what gets sent changes on its own.

Where a Person Decides

Step 2, Code Assigning. Sign the day's codes, or send back the ones you disagree with. Pick between the two codes on the 31 the pipeline would not order. a coder signs the code.
Step 4, Claim Releasing. Read six lines and make six calls. Open the chart only when the line is not enough. a person sends the claim.
Step 5, Denial Working. Approve the four drafted appeals, or cut the list. Read the three denials nobody can explain. an appeal costs time you own.
Step 6, Payer Learning. Accept or reject the medical necessity rule. Take the coding habit back to the facility it belongs to. nobody changes what we send but you.

Operating Model

This changes how work flows through the team.

Role
Responsibility
Code Assigning owner
Sign the day's codes, or send back the ones you disagree with. Pick between the two codes on the 31 the pipeline would not order. a coder signs the code.
Claim Releasing owner
Read six lines and make six calls. Open the chart only when the line is not enough. a person sends the claim.
Denial Working owner
Approve the four drafted appeals, or cut the list. Read the three denials nobody can explain. an appeal costs time you own.
Payer Learning owner
Accept or reject the medical necessity rule. Take the coding habit back to the facility it belongs to. nobody changes what we send but you.
Claim Checking, when it goes wrong
Look at the blocked authorizations if the queue is growing.

What Transfers, What Must Be True

What transfers
A person is in the loop wherever a clinical judgement becomes a billing claim. Four of the six stations refuse to proceed on their own, Code Assigning, Claim Releasing, Denial Working and Payer Learning, and those are the four a person carries. Everywhere else the subagents run at volume and reach you only when they cannot settle something.
Code Assigning stops for a person, and a coder signs the code.
Claim Releasing stops for a person, and a person sends the claim.
Denial Working stops for a person, and an appeal costs time you own.
Payer Learning stops for a person, and nobody changes what we send but you.
Every subagent says what it will not do. 13 of them do.
What must be true in your environment
The agent can read the systems your records already live in. This one reads 30.
Somebody owns Code Assigning and has time for it.
Somebody owns Claim Releasing and has time for it.
Somebody owns Denial Working and has time for it.
Somebody owns Payer Learning and has time for it.

Failure Modes

What breaks this pattern:

✗ Charges the note does not support

Nothing holds a charge when the note falls short. The claim pays, an audit later finds the gap, and the practice refunds the money with penalties.

✗ Codes nobody signed

A suggestion becomes a billed code without a certified coder's name on it. When a payer questions the code, no person can say why it was chosen.

✗ Claims sent without prior authorization

A claim goes out for a service that needed authorization first. The payer denies it, and the money is gone, because authorization cannot be granted after the fact.

✗ Rules built on one denial

The system writes a blocking rule from a single denial. One payer mistake becomes a standing rule, and clean claims get held for a reason that was never real.

Directional Outcomes

What the agent counts, and the station that counts it.

These counts are the tallies from one monitored run of the agents. They are not monthly or annual totals.

Encounters stamped
Counted at Charge Reading
6,800
Coded off an unread note
Counted at Charge Reading
0
Held, note short of the level
Counted at Charge Reading
63
Claims checked
Counted at Claim Checking
6,779
Sent with a known edit failing
Counted at Claim Checking
0
Held for a person
Counted at Claim Checking
34
Our measurement policy: We do not publish precise ROI without baseline methodology. Every figure above carries its basis.

What Runs Where

Every step names the subagent that does the work, the record it writes, the thing that raises a question for a person, and what it is allowed to touch. This is drawn from the source, not from a diagram somebody kept in sync by hand.

1Charge Reading
subagentcharge-read
writesclaims/captured/<enc-id>.json
raisesnote-does-not-support
may touchclaims/**, ehr/** read-only
2Code Assigning
GATE
subagentcode-suggest
writesclaims/coded/<enc-id>.json
raisesawait-coder-signature
may touchclaims/coded/**, ehr/** read-only
3Claim Checking
subagentpre-submit
writesclaims/checked/<claim-id>.json
raisesmissing-prior-authorization
may touchclaims/**, payers/** read-only
4Claim Releasing
GATE
subagenthold-brief
writesclaims/decisions/<claim-id>.json
raisesawait-human-approval
may touchclaims/decisions/**, everything else read-only
5Denial Working
GATE
subagentdenial-work
writesclaims/denials/<claim-id>.json
raisespropose-appeal
may touchclaims/denials/**, ehr/** read-only
6Payer Learning
GATE
subagentpayer-read
writesanalytics/payers/<period>.json
raisespropose-rule-change
may touchanalytics/**, payers/** read-only

Stack

Every system this agent reads or writes.

System
Read at
Stations
the EHR clinical notes
Charge Reading
1 of 6
the EHR documentation
Denial Working
1 of 6
the EHR note behind it
Claim Releasing
1 of 6
the EHR notes
Code Assigning
1 of 6
the appeal history
Denial Working
1 of 6
the authorization records
Claim Checking
1 of 6
the authorization system
Claim Releasing
1 of 6
the charge master
Charge Reading
1 of 6
the claim record
Claim Releasing
1 of 6
the claim scrubber
Claim Checking
1 of 6
the clearinghouse queue
Claim Releasing
1 of 6
the clearinghouse rejections
Denial Working
1 of 6
the code sets in force on the service date
Code Assigning
1 of 6
the coder work queue
Code Assigning
1 of 6
the contract terms
Payer Learning
1 of 6
the denial history
Payer Learning
1 of 6
the eligibility service
Claim Checking
1 of 6
the facility schedule
Charge Reading
1 of 6
the first-pass rate by facility
Payer Learning
1 of 6
the payer bulletins
Payer Learning
1 of 6
the payer coding policies
Code Assigning
1 of 6
the payer denial reason codes
Denial Working
1 of 6
the payer edit libraries
Claim Checking
1 of 6
the payer edit that fired
Claim Releasing
1 of 6
the practice management encounters
Charge Reading
1 of 6
the provider roster
Charge Reading
1 of 6
the provider's own coding history
Code Assigning
1 of 6
the referral records
Claim Checking
1 of 6
the remittance advice
Denial Working
1 of 6
the validation rule table
Payer Learning
1 of 6
Next Step

Want to see if this pattern fits your healthcare claims?

No build commitment·Real samples, not a demo·Estimate in writing