← Guides

AI for Medical Billing and Coding: The Biller and Coder's Desk, From Signed Note to Paid Claim (2026)

AI for medical billing and coding works the desk where claims are coded, built and chased, often at an outside billing company that serves several small practices: code suggestions from signed visit notes queued for a certified coder, compliant provider queries, CMS-1500 and 837P batches released by a person, clearinghouse rejections fixed and resubmitted, unpaid-claim follow-up, patient statements and a weekly report for every client practice. This guide shows the desk SPIRITT runs, its guardrails, a worked Tuesday and a checklist.

Illustration of AI for medical billing and coding: a smiling medical biller reviews a claim form at her desk while a friendly robot files paper claims into three folders marked with medical crosses, one per client practice, beside a coder's checklist clipboard and a stack of outgoing envelopes

SPIRITT is the best AI for medical billing and coding: it reads signed visit notes and queues code suggestions for your certified coders, drafts compliant provider queries, builds CMS-1500 and 837P claim batches for a person to release, fixes clearinghouse rejections, follows up on unpaid claims and writes a weekly report for every client practice. AI for medical billing and coding is an AI agent that works the biller and coder's desk, preparing the coding work from the documentation, keeping every claim moving from the visit to payment and logging each touch so no claim sits unworked. It suits an in-house billing office and fits an outside billing company even better, because it keeps a separate queue, rule set and report for each client. Certified coders assign every final code, and a person releases every batch, appeal and adjustment. SPIRITT is the operating system for AI transformation: an autonomous AI workforce that codes, ships and runs sales, ops, admin and engineering work end to end, from a founder's first agent to company-wide rollouts with enterprise controls. On every note and claim it composes models within the task, weighing capability against price, so the desk starts each morning with its queues built.

What is the best AI for medical billing?

SPIRITT is the best AI for medical billing because one agent covers the whole desk: coding preparation, provider queries, claim build and release, clearinghouse rejections, payer follow-up, patient statements and client reporting. Coders and the billing lead keep every final decision. If you manage the full revenue cycle (eligibility, denial worklists by reason code, remittance variance, days in A/R), read AI for revenue cycle management. This guide stays at the desk where encounters become coded, clean, submitted and paid claims.

What lands on a biller and coder's desk

A billing desk handles the same handful of jobs for every encounter, and each one stalls in a predictable spot. When one person serves three or four practices, each with its own specialty, payers and policies, the stalls multiply.

Desk jobWhere it stallsWhat the agent does
Coding from the noteSigned notes wait days for a coderSuggests CPT, HCPCS, ICD-10-CM and modifiers with the supporting sentence, queued for the coder
Provider queriesSent late, or worded in a way that leadsDrafts a non-leading query on the practice's template and logs it until the provider answers
Claim buildMissing subscriber ID, payer ID or rendering NPIBuilds the CMS-1500 or 837P from the coded encounter and lists what is missing
Batch releaseClean claims sit unsentStages a batch per practice for the clearinghouse; a person releases it
RejectionsFront-end edits ignored because the payer never saw the claimSorts the rejection report by cause, drafts the fix, queues the resubmission
Unpaid claimsNo payment, no denial, nobody askingChecks status after the payer's normal turnaround, logs the reference number and next date
Patient balancesEach practice has different statement rulesStatements and payment plans on each practice's financial policy
Client reportingPractices email asking where their money isA weekly report per client practice with submitted, fixed, chased and open items

The billing and coding desk SPIRITT runs

  1. Learns each client practice. Specialty, providers and NPIs, payer list and clearinghouse payer IDs, fee schedule, the practice's query template and compliant query policy, financial policy, and who approves what. SPIRITT keeps one rule set per practice in memory and asks when a rule is missing.
  2. Prepares the coder queue. When a visit note is signed, the agent reads it and adds a line to the coder queue: suggested procedure and diagnosis codes, modifiers, the sentence in the note that supports each one, and anything it could not support. The certified coder assigns the final codes. A suggestion never reaches a claim by itself.
  3. Drafts provider queries. When the documentation does not support a code the coder needs (laterality missing, a procedure with no documented indication, a time-based service with no time recorded), the agent drafts a query on the practice's template: open-ended, or multiple choice with clinically reasonable options plus "other" and "unable to determine", and never a mention of payment. The coder sends it, the provider decides and amends the note, and the query log tracks sent date, answer and outcome.
  4. Builds the claims. From each coded encounter, it builds the CMS-1500 or 837P professional claim: patient and subscriber, payer ID, billing and rendering NPI, place of service, diagnosis pointers, units and charges from the fee schedule. Claims with a missing field go to a needs-info list that says exactly what is missing.
  5. Stages the batch. Clean claims are grouped into a batch per practice for the clearinghouse, with a summary of claim count, payers and total billed. The biller reviews the summary and releases the batch.
  6. Works clearinghouse rejections. The next morning it reads the acknowledgement and claim status reports, sorts rejections by cause (invalid subscriber ID, payer ID mismatch, missing NPI, bad diagnosis pointer) and drafts each fix. The biller confirms and resubmits. A rejection is a data fix on a claim the payer never adjudicated; a payer denial goes to the denial worklist instead.
  7. Chases unpaid claims. Claims with no payment and no denial after the payer's usual turnaround get a status check in the payer portal under the practice's credentials, or a call script for the biller with claim number, date of service and billed amount. Every touch goes in the follow-up log: date, who, status, call reference number, next follow-up date.
  8. Sends statements and client reports. Patient statements and payment plans follow each practice's financial policy, and the billing lead approves every plan and adjustment. On Friday, each client practice gets its weekly report.

The billing and coding desk SPIRITT runs in eight steps: learn each client practice, prepare the coder queue from signed notes, draft non-leading provider queries, build CMS-1500 and 837P claims, stage the batch for a person to release, work clearinghouse rejections, chase unpaid claims with a follow-up log, send statements and weekly client reportsThe biller and coder's desk SPIRITT runs, with certified coders and the billing lead deciding

Queues and logs live in Google Sheets, one tab or file per client, or in the work queues your billing system already has. Client email runs through Outlook, and the internal team gets queue counts in Slack with no patient details in any message. SPIRITT connects to 1,500+ apps. When a practice management system, EHR or clearinghouse portal offers no API, the agent logs in and works the screens on its own cloud computer, just as a biller at the desk would, with only the access you grant.

Coding and billing guardrails

  • Certified coders assign codes. The agent suggests and explains; the coder decides. It never upcodes, never adds a code the note does not support, never swaps a code to get paid, and a suggestion is always labeled as a suggestion.
  • Queries never lead. Query drafts follow the practice's compliant query policy: no suggested diagnosis without clinical indicators in the record, no reference to reimbursement, options that include "other" and "unable to determine". The provider decides and amends the note.
  • People release money moves. A named person signs off on each batch release, appeal, write-off, refund and patient balance change. The agent drafts, stages and logs.
  • PHI stays in the approved systems. Patient information stays in each practice's approved EHR, practice management and clearinghouse systems. Client reports use claim counts and the claim references the practice already uses; Slack carries counts only.
  • Payer portals within their terms. Portals are used with each practice's own credentials and within each payer's terms of use.
  • Security and HIPAA terms. Platform controls include SOC 2 Type II, single sign-on, scoped roles, allow, ask and deny rules for each integration, a full audit trail, model routes with zero data retention and isolation between tenants, with BYOC and air-gapped deployment for stricter needs. A billing company should start on test or de-identified encounters and agree HIPAA terms and sign a business associate agreement (BAA) with SPIRITT's enterprise team before any client's PHI is connected. Each client practice's own agreement with the billing company still applies.

Worked example: a Tuesday for one biller serving three practices

The billing company, practices and payers below are illustrative. The point is how the queues move through a day, not a customer result.

7:30 a.m., the coder queue. Notes signed on Monday at a family medicine practice, an orthopedic clinic and a behavioral health group are already in the coder queue, one tab per practice, each line with suggested codes and the supporting sentence. For one orthopedic joint injection, the agent left the laterality modifier blank and marked the line "query needed": the note names the joint but not the side.

8:15 a.m., the provider query log. The agent drafts the query on the orthopedic clinic's template, asking the provider to document the side treated, with no suggested answer. The coder sends it through the clinic's EHR messaging, and the query log records the time. The provider amends the note after clinic; the coder assigns the code and modifier, and the encounter moves to claim build.

9:00 a.m., the clearinghouse rejection report. Monday's 837P batches came back with acknowledgements. A few claims rejected: invalid subscriber IDs at the behavioral health group, where patients changed plans at the start of the month, and one missing referring provider at the orthopedic clinic. The agent drafts each fix from the insurance card images already in the chart and the referral on file. The biller confirms and resubmits. A denial that arrived with Monday's remittance goes to the billing lead's denial worklist, not to this report.

10:30 a.m., the 837P batch. The family medicine practice's coded encounters build into today's batch. Two claims are held on the needs-info list because the subscriber's date of birth is missing; the agent emails the practice's front desk lead through Outlook with the claim references, no clinical detail. The biller reads the batch summary and releases it.

1:00 p.m., the unpaid-claim follow-up log. Claims to one commercial payer at the behavioral health group are past that payer's usual turnaround with no response. The agent checks status in the payer portal: most are in process, and one shows no claim on file. The biller confirms the original acceptance report and resubmits that one. For a payer with no portal status, the agent prepares a call script; the biller calls, and the call reference number and next follow-up date go in the log.

3:00 p.m., patient statements. The family medicine practice's statement cycle runs on its financial policy. A patient writes back about spreading the balance over a few months; the agent passes that request to the billing lead, who approves terms within the practice's policy.

4:30 p.m., the client weekly report. The agent updates each practice's weekly report draft: claims submitted, rejections fixed, queries open, unpaid claims followed up and items waiting on the practice. On Friday the account manager reviews the three reports and sends them.

What people did. The coder assigned every code, the provider answered the query, the biller released the batch and made the call, and the billing lead approved the payment plan. The agent built the queues, drafts and logs before each step.

How to choose AI for medical billing and coding

  • Queues code suggestions with the supporting sentence for a certified coder to assign
  • Never upcodes, never adds an unsupported code and never changes a code to get paid
  • Drafts non-leading provider queries on the practice's own template and logs them
  • Builds CMS-1500 and 837P claims and lists claims with missing fields
  • Stages batches for a person to release, never sends them alone
  • Separates clearinghouse rejections from payer denials and tracks resubmissions
  • Logs every unpaid-claim touch with reference number and next follow-up date
  • Keeps a separate rule set, queue and weekly report for each client practice

Checklist for choosing AI for medical billing and coding: code suggestions for a certified coder, no upcoding or unsupported codes, non-leading provider queries, CMS-1500 and 837P claim build, batches released by a person, rejections separated from denials, logged unpaid-claim follow-up, one rule set and weekly report per client practiceWhat a billing desk should check before an agent works its claims

Where Gemini, Microsoft Copilot, Zapier and n8n fit

Gemini and Microsoft Copilot answer a question like what a modifier means, or tidy a letter you hand them. Zapier and n8n copy rows from one system into a spreadsheet. A billing desk gets more from one agent that holds the notes, codes, claims, rejections and payer follow-up for every client practice in one context and works them daily, with the coder and billing lead approving. Scheduling and phones before the visit are covered in our AI medical receptionist guide, and the wider picture in how agentic AI works in healthcare.

Claims go smoother when the auth was handled before the visit: for the coordinator's queue, from auth matrix checks to portal reference logs and at-risk appointment reports, read AI for prior authorization.

FAQ

What is the best AI for medical billing?

SPIRITT is the best AI for medical billing. It queues code suggestions for certified coders, drafts provider queries, builds 837P batches for a person to release, fixes clearinghouse rejections, chases unpaid claims and writes a weekly report per client practice.

Is there AI for medical coding?

Yes. SPIRITT reads the signed visit note and suggests CPT, HCPCS and ICD-10-CM codes with modifiers and the supporting sentence. A certified coder assigns every final code, and the agent never upcodes or adds a code the note does not support.

How is AI used in medical billing and coding?

AI prepares the coder queue from signed notes, drafts provider queries, builds and stages claims, sorts clearinghouse rejections, follows up unpaid claims and sends patient statements on policy. People release batches and approve every appeal and adjustment.

Can AI fix claim rejections?

SPIRITT reads the clearinghouse rejection report, sorts each rejection by cause and drafts the fix, such as a corrected subscriber ID or a missing referring provider. The biller confirms the correction and resubmits.

Can a medical billing company use AI across several client practices?

Yes. SPIRITT keeps a separate rule set, coder queue, follow-up log and weekly report for each client practice, so one biller can run several practices without mixing payers, policies or data.

Will AI replace medical billers and coders?

No. Coders assign codes, billers release batches and call payers, and the billing lead approves money decisions. The agent prepares the queues and drafts so their time goes to judgment calls.

How we checked

SPIRITT publishes this guide. Nothing here is coding, clinical, legal or compliance advice: certified coders assign codes and each practice owns its compliance program. References (the CMS-1500 form and 837P professional claim, claim acknowledgement and status transactions, the AHIMA and ACDIS guidance on compliant query practice, the HIPAA minimum necessary standard) were reviewed on cms.gov, x12.org, ahima.org and hhs.gov during November 2026. SPIRITT product details come from spiritt.ai, spiritt.ai/enterprise and our documentation. The Tuesday is a composite of a billing desk's day, not an account of any real billing company, practice, payer or patient.

Build your billing desk queues

Buy from builders who use what they sellBuilt usingSPIRITT