AI for US medical billing
ClaimVise checks eligibility, pulls charts, codes claims and posts charges. Your people review every claim and click Post. It is made for US medical practices and the billing teams that serve them. Proven in hematology and oncology, with a rollout path for other specialties.
Live with a billing-agency client: more than 1,200 eligibility checks and 250 claims so far (October 2026). See what is in place and what is planned.
Human oversight
ClaimVise supports your billing team. It does not replace it. Here is where a person stays in control.
Nothing goes into a practice's billing system until a reviewer approves it. You can also export an 837P file instead.
The connector checks that a note is signed. It never codes from a draft.
If a connector hits a screen it does not recognize, it stops and explains what happened and who should fix it. It does not guess, and it never retries a failed row on its own.
The AI suggests codes and flags risks. Supervisors and coders review the exceptions. Every override goes in the audit log.
See how PHI is protected, and what is still planned, on the HIPAA page.
The AI agents
Each agent has one job. They read the note, code the claim, check it against the rules, score the denial risk and log what they did. A supervisor reviews every claim before it is exported or posted, and the agents flag the ones that need a closer look.
Reads clinical notes in any format.
Typed notes, PDF superbills, handwritten or scanned charts, and OncoEMR charts with labs and medication administration records (MAR). It pulls out the diagnoses, procedures, time spent and infusion details the Coding Agent needs.
Assigns ICD-10 and CPT codes, with a confidence score.
Codes office visits, procedures, preventive visits and infusion across specialties. A fast model reads the note first. A stronger model then codes it against current CMS data and any specialty rules set for the practice. It is proven in depth in hematology and oncology. For any other specialty, its rules run in shadow on your own notes before routine use.
Checks every claim against fixed rules.
About 2.6M NCCI procedure pairs from the current CMS release, MUE unit limits, CMS modifier rules (JZ, JW, KP, KQ, Q0, Q1, XU), ICD-10-CM validity and HCPCS codes, plus payer rules set per agency, practice and payer. CMS data is refreshed every quarter. A claim with a blocking finding stays on hold until a person clears it.
Scores denial risk before you submit.
Fixed rules score known problems first. Then an AI pass adjusts the score. Each claim lands in a low, medium, high or critical lane, so reviewers know where to look first. It is a guide, not a promise of payment.
Drafts appeal letters from the original clinical note.
Pulls the diagnosis, the documented procedure and the medical necessity language from the note. It drafts a letter with clinical citations for a supervisor to review.
Checks every drug billing unit against CMS and FDA data.
Covers more than 135 drugs, including oncology infusions, IVIG, biologics and biosimilars. It reconciles multi-drug visits from the MAR, handles waste and multi-vial modifiers, applies practice-supplied (IOD) drug rules and matches NDCs for charge posting. If a unit cannot be backed by reliable data, it does not guess. It flags the line for a person.
Logs every PHI access and every agent decision.
Every event gets a SHA-256 integrity hash that is checked each time it is read, supporting HIPAA §164.312(b) audit controls. Any change to a log entry can be detected. Every claim change, override, approval, export and posting request is recorded with who, when and what happened.
Secure connectors
Most billing time is not coding. It is logging in to portals, pulling charts and re-typing charges. ClaimVise connectors do that work inside the practice's own EMR, billing system and eligibility portal, one session at a time. The logins sit in an encrypted vault that the agency controls. A connector never retries a failed row on its own. A person decides.
Pulls the signed chart for every billed visit.
Upload the practice's daily signed-charges export. The connector finds each patient in OncoEMR, opens that visit's documents for that date and gathers the billed items: lab results, the medication administration record and the visit note. It merges them into one PDF for the Coding Agent. If one patient fails, the session recovers and the next patient is not affected.
Posts reviewed charges into the practice's billing system.
After a reviewer clicks Post, the connector opens the right batch for the date of service, finds the patient and enters each charge line with provider, facility, authorization and diagnosis. Then it checks the result. Drug lines come from the practice's own dispense record and NDC list, and a drug dispensed under several NDCs posts as one line per NDC. Every attempt keeps screenshots, and a retry never re-posts a group that already went in. A plain-words readiness report shows missing dispense or NDC data before claims reach the connector.
Checks eligibility straight from the practice's worklist.
Upload the eligibility worklist. The connector signs in to the eTactics portal (a person types the one-time code when the portal asks), checks each patient one at a time and fills in status, effective date, co-pay, deductible, co-insurance, out-of-pocket maximum, plan type and remarks in a verified copy of the same workbook. Medicare and commercial payers follow different steps, set per practice.
Live today: OncoEMR, AdvancedMD and eTactics. Notes from any other EHR can be uploaded as a PDF or image. Planned, not yet live: Availity, eClinicalWorks, OmniMD and Meditab (see the roadmap).
The problem
Billing agencies and in-house teams are asked to do more for each client at a lower price per claim. Payer rules, drug billing and prior authorization keep getting more detailed. Most of the work is not clinical judgment. It is repetitive work that follows rules.
Eligibility checks, chart pulls and charge entry mean logging in to three or four systems for every patient and re-typing what is already on screen. It is slow, and one typo can become a denial.
Infusion practices bill by the milligram, with waste, multi-vial and practice-supplied drug rules, and a modifier for each. One wrong unit or code means a rejected line or an overbill.
An agency with many practices needs to see, by practice and by person, what was pulled, coded, reviewed, approved and posted. And it needs to prove it later.
Fixed headcount costs too much in a slow month and falls short in a busy one. Software work should be priced and capped per practice, with extra charges only when volume really goes up.
How it works
Agents and connectors do the repetitive work. A person reviews every claim before it is exported or posted. Nothing is posted until someone clicks Post. Then ClaimVise learns from what your reviewers changed.
The eTactics connector works through the eligibility worklist and returns a verified copy with the benefits filled in.
The OncoEMR connector pulls each billed visit's signed documents. Or the Scribe Agent reads an uploaded note, superbill or scanned chart.
The Coding Agent assigns ICD-10 and CPT codes with a confidence score. If confidence is low, a stronger model takes over.
Scrubbing, drug checks and denial scoring look at units, modifiers, bundling and payer rules. Blocking findings hold the claim.
A supervisor reviews the claim, edits lines if needed and approves. The original AI output is always kept next to the human change.
The AdvancedMD connector posts the charges, or you export an 837P file. If a claim is denied, the Appeal Agent drafts the letter.
Measure and learn
Reviewers end up fixing the same things again and again. ClaimVise notices, checks it against what the practice really charged, and turns the strong patterns into rules you can see and switch off.
Every night the billing scorecard compares each claim three ways: as ClaimVise coded it, as your reviewer finalized it, and as the practice actually charged it. You get trends over time, each reviewer's edits and a list of the claims that differ. We switch it on for your agency, and your admins see it.
When reviewers keep making the same correction and the practice's own charges agree, ClaimVise proposes a rule for that practice. A rule can only remove a code or add a modifier from a short approved list. It never makes up a billed code.
Rules are off until you turn them on for a practice. You can see every rule, pause it, and spot it on the claim. A rule stops itself if reviewers start undoing it, and it never touches a line a person has corrected.
The AI model itself is not retrained on your data. ClaimVise learns rules, and you decide which ones stay. We do not publish an accuracy percentage. The scorecard measures yours.
Specialties
ClaimVise was not built for one specialty. The coding engine and the CMS reference data cover them all, and the platform keeps a list of more than 100 specialties and subspecialties. Hematology and oncology is the one running on real claims today. Each new specialty gets its own rules and is tested beside the standard result before it is used for billing.
Infusion and injectable drug billing is where most errors hide: units by the milligram, waste and multi-vial rules, practice-supplied drugs, biosimilars, chemo administration codes and NDC matching. This is where ClaimVise is tested hardest, and it is live with a billing-agency client.
Office visits, procedures and preventive visits go through the same Coding Agent. NCCI and MUE edits, the fee schedule, ICD-10-CM and HCPCS are loaded for all specialties and refreshed each CMS quarter. Payer rules are set per agency, practice and payer.
Each practice is set to its specialty. A new specialty gets its own rules and runs in shadow on your real notes first. A qualified coder must review and sign off the rules before routine use. A person still reviews every claim before it is exported or posted.
What the list covers
The list follows the Medicare specialty codes, with pediatric subspecialties and urgent care added. A specialty on the list is one we can set up. It does not mean it is tested.
General and family practice, internal medicine, pediatrics, geriatrics, preventive medicine and hospitalists.
Cardiology, gastroenterology, neurology, nephrology, pulmonary, endocrinology, infectious disease, rheumatology, allergy and immunology, and more.
General, orthopedic, neurological, cardiac, vascular, plastic and colorectal surgery, urology, otolaryngology and more.
Hematology and oncology, plus surgical and radiation oncology and cell therapy.
Obstetrics and gynecology, gynecologic oncology and midwifery.
Sixteen pediatric subspecialty certificates, from adolescent medicine to pediatric transplant hepatology.
Emergency medicine, urgent care and intensive care.
Diagnostic and interventional radiology, nuclear medicine, pathology and sleep medicine.
Anesthesiology, nurse anesthetists, and interventional and general pain management.
Psychiatry, psychology, counseling, social work and addiction medicine.
Physical and occupational therapy, speech and hearing, chiropractic, podiatry, optometry and nutrition.
Nurse practitioners, physician assistants and clinical nurse specialists.
How it works
Each practice is set to its specialty from a list of more than 100 specialties and subspecialties. The list shows what we can set up, not what is already tested.
A qualified coder must review and sign off the rules for the specialty. Rules nobody has signed off can only run in shadow.
The specialty result is recorded next to the standard result on your real notes. It never changes what is billed. If the specialty rules fail, the standard result is used.
When the shadow results are measured with you and a qualified coder has signed off the rules, the practice moves to routine use. A person reviews every claim before it is posted or exported.
This shows the process. Hematology and oncology is the only specialty running on real claims today.
Platform depth
Under the agents and connectors is what lets a billing agency run many practices on one system: separate practices, settings at every level, current CMS data, and checks that refuse to make up a number.
Each agency sees only its own practices. Each practice has its own plan, prepaid wallet, module access, providers, locations and billing rules. Users get a role (agency admin, supervisor, coder or AR caller), and a practice switcher keeps screens, dashboards and the claims queue on the practice you are working in.
A team report shows, for each person, how many charts they pulled, coded, reviewed, approved and posted.
System, agency, practice, then payer. Agencies bill differently, practices within an agency differ, and payers reject different things. ClaimVise combines all four layers into one billing decision per claim. Agency and practice admins set their billing rules on their own screens.
Example: Anthem rejects KP and KQ modifiers, so the payer layer bills Anthem claims as combined vials while the same practice's Medicare claims still use KP and KQ.
NCCI pairs and MUE limits, Medicare fee schedule RVUs, ASP drug payment limits, HCPCS Level II and ICD-10-CM, plus a dictionary of more than 135 drugs and biosimilars built from CMS and FDA labeling. Each quarterly refresh loads as a versioned run that can be rolled back, and the platform shows how current each dataset is. The October 2026 CMS release, including ICD-10 for fiscal year 2027, is loaded.
If a unit or code cannot be verified, it is left empty and flagged. It is never estimated.
A charge comes only from the practice's contracted rate for that payer. With no reliable rate, the line is marked unresolved and the export is blocked. It is never filled in with a Medicare rate, zero or a default. The same rule covers place of service and date of service.
When a connector cannot continue, the screen says what went wrong and who should fix it, in plain language.
How we build
Healthcare billing is no place for ship now, fix later. Every change to the coding pipeline is tested against real clinical notes before it ships. Every decision is logged. Every connector action leaves a record.
Changes to the coding agent, validators or scrubbing rules must pass the full automated test suite and a run against real oncology infusion notes with medication records before they go live. Changes to connectors are checked against the real portal pages the connector saw.
The code that applies CMS modifiers (JZ, JW, KP, KQ, Q0, Q1, XU) is locked with a hash, so any change fails three separate tests. Every audit event carries a SHA-256 hash that is checked again each time it is read.
Our AWS Business Associate Addendum was signed on May 22, 2026. All AI processing runs on AWS Bedrock under it, and the platform will not start unless the AI provider is set explicitly. Connector logins live in an encrypted vault that agency admins manage. Logins are rate-limited, uploads are checked by file signature, and each agency and practice is isolated. The full move to AWS hosting is targeted for November 30, 2026.
If a billing unit cannot be verified, the drug line is flagged, not billed. If confidence is low, the case goes to a stronger model. If there is no contracted rate, the export is blocked. If a connector does not recognize a screen, it stops and tells a person. These checks help reviewers, and a person always makes the final call.
How we compare
ClaimVise is not a faster coding floor, and it is not a workflow tool with an AI assistant added. It is built for billing agencies. You keep the client relationship. The agents and connectors take the routine work. Your supervisors handle the exceptions.
| Area | Traditional billing operation | Typical RCM software | ClaimVise |
|---|---|---|---|
| Who does the routine work? | Coders and billers, by hand | The same people, with software organizing the queue | AI agents and connectors. A supervisor reviews every claim. |
| Eligibility checks | Manual portal lookups | Often a separate product or clearinghouse feature | A connector works the worklist and returns a verified workbook |
| Chart pulls and charge entry | Manual, patient by patient | Re-typing into the practice system | Connectors pull from OncoEMR and post to AdvancedMD after a person clicks Post |
| Drug billing (infusion) | Depends on the coder | Generic edit rules | Unit, waste, modifier and NDC checks from a sourced drug dictionary |
| Pricing | Per-claim labor or headcount | Subscription plus a share of collections | Per practice per month with volume included, or per unit. Extra volume comes from a prepaid wallet. |
| Many practices, one team | Spreadsheets and shared inboxes | Varies by product | Separate plans, wallets, modules, roles and a team report for each practice |
| Measuring quality | Manual spot checks | Varies by product | A nightly scorecard compares what was coded, reviewed and charged |
| Audit trail | Manual logs | Often partial | SHA-256 checked log of every decision and action |
| Availability | Business hours | Software runs all day, support in business hours | The platform and connectors run around the clock |
Pricing
Pick the model that fits each practice. A unit is one claim (coding) or one eligibility check. Smaller practices usually pay less per unit. Busier practices pay less on a monthly plan with volume included. An agency can mix models across its practices. There are no per-seat fees.
Per practice, per month. Best above about 1,000 claims a month
Per practice, per month. Best above about 2,400 verifications a month
Per claim. Best below about 1,000 claims a month
Per verification. Best below about 2,400 verifications a month
Roadmap
Today ClaimVise covers the front and middle of the revenue cycle: eligibility, chart retrieval, coding, checks, review and charge posting. Next come more integrations, then claim submission and payment follow-up. Items marked planned are not live yet.
eTactics connector for Medicare and commercial payers, with a verified workbook returned per worklist.
OncoEMR connector retrieves the signed documents for every billed visit.
AdvancedMD connector posts reviewed charges after a person clicks Post.
A nightly comparison of what was coded, what reviewers approved and what the practice charged. Switched on per agency.
Rules built from repeated reviewer corrections. Off until you turn them on for a practice.
837P file download. A remittance (835) reader is built and tested, and waits for the first live file.
Moving the full platform onto AWS. Target: 30 November 2026.
Planned: more specialties beyond hematology and oncology, each run in shadow beside the standard result before routine use. No dates yet.
Planned: real-time eligibility and direct claim submission (837) with acknowledgment tracking through Availity.
Read real 835 remittance files, post payments and flag underpayments against contracted rates.
Planned: eClinicalWorks, OmniMD and Meditab, in the order customers ask for them.
Aging reports and payer-specific follow-up steps.
Spot requirements before scheduling, draft the letter of medical necessity and track the approval.
Verify and record secondary coverage on the same worklist.
Find underpaid claims and prepare the supporting data.
Roadmap items are plans, not promises, and can change. Active customers get target dates directly.
FAQ
A tool with an AI assistant still needs people to do the routine work. In ClaimVise, AI agents do that work. Seven agents read the note, suggest codes, check the claim, score denial risk and draft appeals. Three secure connectors work inside the EMR, the billing system and the eligibility portal. A person reviews every claim before it is exported or posted, and nothing posts until someone clicks Post.
Medical billing agencies that run billing for several practices, whether they work from India, the United States or elsewhere. Also in-house billing teams at larger practices. One agency account holds every practice it serves, each with its own plan, wallet, modules and users.
No. The coding engine handles office visits, procedures, preventive visits and infusion, and the CMS data covers every specialty. We started with hematology and oncology because infusion billing is the hardest. That is where we have the deepest testing and a live client. Other specialties are added one at a time, each with its own rules and a shadow run on your own notes first (see the next question). We do not claim measured results outside hematology and oncology yet.
The practice is set to its specialty. The specialty gets its own rules. It first runs in shadow: the specialty result is recorded next to the standard result on your real notes and never changes what is billed. When the shadow results are measured with you and a qualified coder has signed off the rules, the practice moves to routine use. A person reviews every claim before it is posted or exported.
Every claim has a confidence score. If it is low, a stronger model takes over, and fixed checks for units, modifiers, bundling and payer rules run after the AI. A person reviews every claim before export or posting, and the original AI output is kept next to any change. We do not publish one accuracy number, because accuracy depends on the specialty and the documentation. Instead, the billing scorecard measures it on your own claims, comparing what ClaimVise coded, what your reviewers approved and what the practice actually charged.
Yes, in a limited and visible way. When reviewers keep making the same fix, and the practice's own charges agree, ClaimVise can turn that into a rule for that practice. A rule can only remove a code or add a modifier from a short approved list. It never makes up a billed code and never touches a line a person has corrected. Rules are off until you turn them on for a practice. You can see and pause every rule, and a rule stops itself if reviewers start undoing it. The AI model itself is not retrained on your data.
Every night it compares each claim three ways: as ClaimVise coded it, as your reviewer finalized it, and as the practice actually charged it. You see trends over time, how each reviewer's edits compare, a claim-by-claim list and which rules fired. We switch it on for your agency, and your agency admins can see it.
ClaimVise signs a BAA with every customer that processes PHI. All AI processing runs on AWS Bedrock under our executed AWS Business Associate Addendum (May 22, 2026), and PHI is not sent to any provider outside that agreement. Every claim change, override, export and posting request goes in an audit log with a SHA-256 integrity check. Each agency and practice is isolated, connector logins sit in an encrypted vault, and uploads are checked by file signature. The HIPAA page lists what is in place and what is still planned.
It replaces the repetitive work, not the people. You keep the supervisor seat: reviewing flagged claims, clearing holds, clicking Post and managing settings. Most of the time saved comes from not logging in to portals, pulling charts and re-typing charges. Your team spends more of its day on exceptions and client service.
Those products manage the workflow, and people still do the coding and the portal work. ClaimVise adds agents that code and check, and connectors that work inside the EMR, the billing system and the eligibility portal. It is also built for billing agencies, with separate plans, wallets, modules and a team report for each practice.
Live today: OncoEMR for charts, AdvancedMD for charge posting and eTactics for eligibility. Notes from any other EHR can be uploaded as a PDF, image or file import. Planned, not yet live: Availity (eligibility and claim submission), eClinicalWorks, OmniMD and Meditab. The order follows what customers ask for. If you use another system, tell us in the demo.
Coding & Claim Readiness includes all seven agents, the OncoEMR and AdvancedMD connectors, claim review, 837P export, dashboards, the team report, the billing scorecard and the audit trail. Eligibility Verification includes the eTactics connector and the verified worklist. Each comes two ways: a monthly plan per practice with volume included (1,200 claims or 3,000 verifications) and extra volume drawn from the practice’s prepaid wallet (an unused balance expires at the start of each month), or pay as you go per claim or per verification, which usually suits smaller practices. See the pricing section for the figures. A monthly server hosting fee of ₹7,000 / up to $175 per practice is added to the plan price.
It depends on the plan. On a monthly plan, the wallet only pays for volume above what is included. Any unused balance expires at the start of each calendar month (India Standard Time). It is not carried forward and it is not refunded, and the included volume starts fresh each month. On pay as you go, the wallet pays for each claim or check as it is processed, and the balance does not expire. Amounts already used for processing are not refundable. The platform Terms of Service set out the details.
Onboarding is setup, not a big implementation project. You add the practice profile and billing rules, store the practice’s system logins in the encrypted vault and switch on the modules that practice needs. We will give you a realistic timeline for your practices in the demo.
See it live
Book a 30-minute call. Bring a clinical note or an eligibility worklist and watch ClaimVise work through it in front of you.
Book a Demo →