Outsourced Medical Billing or In-House AI Tools? A Decision Rule

Convert both options to a cost per claim on your own volume. Outsourcing usually wins for a one- or two-provider practice with no trained biller, and in-house with AI tools wins when volume is steady and you want control. Published fees run about 4 to 10 percent of collections or $3 to $12 per claim.

Those ranges come from vendor-written pages, so they are a starting point rather than a benchmark. What decides the question is four tests you can run in an afternoon: what each route costs per claim, what you can see and control, what the reports tell you, and how easily you can leave. Run all four on your own figures, because a practice that collects $60 a claim and one that collects $120 can reach opposite answers from the same quotes.

Follow me on Instagram@sagnikteaches

First, what "AI-assisted billing" means on each side

Outsourced billing companies increasingly describe their service as AI-assisted. In practice that usually means automated claim scrubbing, eligibility checks, denial prediction and drafted appeal letters running inside the company's own workflow, with its staff in charge. For example, CareCloud's home page lists cirrusAI Appeals for appeal letters among its AI tools, and its revenue-cycle page calls the billing service "AI-powered" and describes catching errors before claim submission; neither page lists prices or contract terms. Steady Medical Billing's page offers a free audit and a custom quote with no figures, and talks of predictive analytics for denials rather than AI. Prices are quoted individually, which is why you must ask in writing.

Connect on LinkedInSagnik Bhattacharya

In-house AI tools are the same kind of function bought by you: a claim-scrubbing feature in your practice software, an AI note-taker that suggests codes for a clinician to confirm, a drafting assistant for appeal letters and patient statements, an eligibility checker. You own the work and the result. If you keep this route small, one trained biller with good tools can do the work of a larger team, but only if that person is reliable and you have cover for holidays and sickness.

Subscribe on YouTube@sagnikteaches

The AI does not change the question. It changes the numbers on the in-house side, and it gives outsourced vendors a talking point. Ask every vendor exactly which steps are automated, which are done by a person, and which AI output you can see.

Test one: cost per claim on your own numbers

A percentage fee and a per-claim fee look different but can be compared in a short sum. You need three figures from your own records: claims sent per month, the average amount collected per claim, and your current monthly cost of doing billing. Here is a worked example with illustrative figures.

Say a two-provider practice sends 600 claims a month and collects an average of $120 per claim, so $72,000 a month. Three quotes:

On these made-up numbers, outsourcing is cheaper per claim, and the percentage model is the most expensive of the three because the practice collects a lot per claim. Now change one input: if the practice's average collection is $60 per claim, the same 6 percent costs $3.60 per claim, well below the $6.33 per-claim quote and close to the $3.33 flat fee, while the in-house route is unchanged at $8.67. This is why you cannot judge by a headline rate. The percentage model rises with your collections and, in a sense, pays the biller more when your fees go up; the per-claim model rises with volume only; the in-house cost is mostly fixed until volume outgrows one person.

Now add the number the vendor pages do not show: what happens to collections. If a good biller lifts collections by 2 percent over your current process (illustrative), that is $1,440 a month on $72,000. That gain does not appear in any fee quote, and it can outweigh the difference between options. So measure your current denial rate and days in accounts receivable first; each option has to beat that baseline, not just the fee.

Test two: control over what is sent and what is appealed

With in-house billing you see every claim. With outsourcing you see what the vendor chooses to show. Ask these questions and write down the answers:

• Can I see each claim before it goes? After it goes? Can I stop one?
• Who decides whether to appeal a denial, and below what dollar value does the vendor stop chasing?
• Who talks to the payer, and does the vendor use my payer portal logins or its own?
• Where do patient statements and patient calls go, and who answers them?

The last one is easy to miss. A billing company that sends statements and answers patient billing calls represents your practice to patients. Read its scripts. An in-house team with a drafted-by-AI statement template keeps that voice yours.

A before-and-after, illustrative: a practice receives a denial for a missing modifier. Before, the biller reads the denial code in the remittance, looks it up, fixes the claim and resubmits in about 20 minutes. After adding an AI claim-check, the same error is flagged before submission in under a minute, and the denial never occurs. The control gain is real but it depends on the biller reading the flags, which is why the tool sits in your hands rather than a vendor's.

Test three: reporting you can read without a vendor

Reports decide whether you notice a problem in week three or month three. Whatever route you choose, ask for the same ten figures, monthly, in a spreadsheet you can open yourself:

1. Claims submitted and claims paid on first submission
2. Clean-claim rate (accepted without correction)
3. Denial rate by payer and by denial reason
4. Days in accounts receivable, and the share older than 90 days
5. Net collection rate
6. Charges, adjustments and payments by provider
7. Appeals filed and appeals won
8. Patient balances outstanding
9. Credits and refunds owed to payers
10. Fees paid to the vendor, matched to the claims counted

Treat "we will send you a dashboard link" as a promise to check, not an answer. Open it during the sales call and ask for last month's report from a practice of your size with names removed. If you go in-house, an AI assistant can help you build the monthly summary from an exported spreadsheet, but check totals against the source. You can use a prompt like the one below.

Here is last month's claim export (columns: payer, claim_id, billed,
paid, denial_code, date_sent, date_paid). Do not guess any missing values.
1. Give the clean-claim rate, denial rate by payer, and average days from
   sent to paid.
2. List the top five denial codes with counts.
3. Flag any claim older than 60 days with no payment.
Show the formulas or the counts you used so I can check them.

A sample output (illustrative) lists a 91 percent clean-claim rate, denial codes with counts, and 14 claims over 60 days. What you fix: confirm the 91 percent against a manual count of one payer, because totals can be wrong when the file has duplicates. Never act on an AI total you have not spot-checked.

Test four: exit terms, tested before you sign

Most of the pain in billing contracts appears on the way out. Write the exit questions into the first call, not the last.

• Notice: how long, and is it tied to a minimum term?
• Fees after exit: does the vendor keep a percentage on payments that arrive after you leave for claims they submitted? For how long?
• Open accounts receivable: who works the outstanding claims during notice, and who after?
• Data: who owns claims, remittances and patient account history, in what format, and how long to receive the export?
• Access: do payer portal logins and clearinghouse enrolments stay in your name?
• Performance: is there a clean-claim or collection target with a right to leave if it is missed?

Ask a qualified adviser to read the contract; billing involves patient data, so a processing agreement is normally part of it (see minutes, BAA and exit terms to negotiate for how the same questions apply to a voice-agent vendor). Enrolments in your own name matter: if the vendor registers with payers under its details, moving later can take months.

A rule you can apply in ten minutes

Score each of the four tests 1 to 3 for each route, where 3 means better for you. Add them. Illustration for the practice above:

These scores are illustrative, and the practice might still outsource because it has no one to hire. Use the rule to show where your preference comes from rather than to decide automatically. Two overrides matter: if you cannot hire or retain a biller within three months, outsource for now with a short term; if the quoted outsourced price is below the in-house cost per claim and the exit terms are clean, outsource and keep a monthly review. Either way, put the decision date and the review date in your calendar.

Where the AI tools sit in each route

If you outsource, you can still use AI on your side: summarise the monthly report, draft questions for the account manager, check that the figures reconcile with your bank deposits. If you stay in-house, start with the cheapest AI step that targets your biggest loss. For most practices that is a pre-submission claim check, because the cost of one avoidable denial is a rework of the same claim and weeks of delay. Next is eligibility checking before the visit (for the dental version see AI for dental billing: claim checks, statements and follow-up), then statement and appeal drafting. Hold the scribe-suggested-coding step to the last; it touches clinical documentation and needs your clinician's sign-off every time.

For the wider buying questions that apply to every AI tool in a practice, use what to ask before buying any AI tool for a medical practice, and for the sequence of a first rollout, a worked AI implementation plan for a small medical practice. For cost ranges on the clinical side of the tool stack, therapy practice software and AI add-ons shows what the add-ons cost.

Fees that hide outside the headline rate

Whatever the model, ask for a list of every other charge. Billing quotes often grow once these appear:

• Set-up or onboarding fees for loading your payers and fee schedules.
• Statement and postage fees for patient bills, sometimes per page.
• Clearinghouse and portal fees, which may be passed through at cost or marked up.
• Credentialing and payer enrolment, usually charged per payer or per provider.
• Minimum monthly fees that apply when your volume is low, for example in a holiday month.
• Charges for old receivables, including a higher percentage for claims older than 90 days.
• Software seats if the vendor uses your practice system and you pay for the extra user.

Add each to your cost-per-claim sum. Illustration: a $2,000 flat fee that is really $2,000 plus $250 statements plus $300 clearinghouse plus a $150 minimum top-up becomes $2,700 for 600 claims, which is $4.50 a claim, not $3.33. Still cheap in this made-up case, but a different ranking against the $6 per-claim quote.

Three practices, three different answers

A solo practitioner with no admin staff. Billing is a few hundred claims a month and the owner does it at night. Outsourcing on a per-claim or flat fee removes the biggest single-point-of-failure risk, and the practice can add an AI statement-drafting tool of its own later. A short contract with a clean exit is more important than the lowest rate.

A four-provider group with a front-office manager. Volume is steady and the manager knows the payers. An in-house biller using AI claim checks, plus one outsourced service only for old receivables over 90 days, often gives the best control and reasonable cost. The hybrid is worth pricing: ask each vendor on the shortlist whether it will work aged accounts only, and on what fee.

A practice that has lost its biller. Cash flow is the emergency. Outsource immediately for a defined period, ask for a 90-day term with a right to leave, and use that time to hire or train. Do not sign a multi-year contract in a crisis.

Finding the break-even claim volume

One more sum helps when volume is changing. Say the in-house route costs $5,200 a month fixed and a per-claim outsourced quote is $6.33 all in. The break-even is $5,200 divided by $6.33, about 821 claims a month. Below that the vendor is cheaper on fees; above it, in-house is cheaper, provided one biller can really handle the volume. If the practice is at 600 claims and growing by about 40 a month, in-house passes break-even in roughly six months. Then decide whether to start in-house now with extra capacity, or outsource now and revisit at 800 claims. Put the review date in your calendar so the decision does not drift.

Remember that one biller has a ceiling too. If your own biller tells you the practice is near what one person can handle, a practice that grows much further needs a second person, and the fixed cost steps up. Ask your biller, not the vendor, what volume they can handle with the tools in place.

An email to send to every billing company on your shortlist

Subject: Written quote request: billing for [number] providers

Hello [first name],

We are a [specialty] practice with [number] providers, about [number]
claims a month and [amount] average collected per claim. Please send in
writing:

1. Your fee model, and the exact figure it applies to (gross charges, net
   collections, or per claim).
2. Every other fee: set-up, statements, clearinghouse, credentialing,
   minimums, aged-account rates.
3. Which steps are automated (claim checks, eligibility, appeals) and
   which a named person does.
4. The monthly reports you provide, with a sample with names removed.
5. Contract term, notice period, fees after exit, data return format and
   timing, and whether payer enrolments stay in our name.
6. Who answers patient billing calls.

Thank you,
[your name]

Compare the replies in one table. A vendor that cannot answer question 1 in a single sentence is telling you something about how its invoices will read.

A realistic mistake: the percentage that applied to the wrong number

Illustrative. A practice signs at "5 percent" and assumes it is a percentage of money received. The contract says "of gross charges billed". The practice's billed charges are well above what payers actually pay, so the effective rate is far above 5 percent of collections. The error surfaced only after the first quarter, when the invoice was compared with the bank deposits. The fix is a single line in the quote: "percentage of net collections received, calculated monthly, shown on the invoice with the claim list". Ask for it every time.

Questions for your current biller before you decide

If you already have someone doing the work, they hold the best data you will get. Ask them, kindly and specifically, which denials come back most often, which payers take longest, how many hours a week go on chasing, and what they would automate first. Their answers show whether AI claim checks would remove real work or only decorate it. They also tell you the cost of switching: a biller who knows your payers' habits is worth keeping, and an outsourced vendor needs months to learn them. If the biller is leaving, record their routines now, in a checklist, because that document makes either route safer, and it is a good first job for an AI drafting assistant working from their notes.

A 30-day trial plan for whichever route you pick

1. Week 1: record your baseline: clean-claim rate, denial rate, days in accounts receivable, monthly billing cost.
2. Week 2: start the new route on one payer or one provider, not all of them.
3. Week 3: compare ten claims in detail against your old process: time spent, errors, payer response.
4. Week 4: read the first report with the vendor or biller, line by line, and agree the monthly review.

Keep the old process alive for one payer until the new route has completed a full payment cycle. If you want to work through the numbers with someone, that is what my AI implementation consultation covers.

Medical billing decision questions

What do outsourced billing companies usually charge?

Fee guides written by billing and outsourcing companies put typical rates at about 4 to 10 percent of net collections, $3 to $12 per claim, or flat monthly fees from roughly $500 to $2,500. Because sellers wrote them, treat them as directional and get written quotes, with the figure the percentage applies to stated.

Does AI replace a medical biller?

Not for a practice that wants control of its denials. AI tools can check claims before submission, draft appeal letters and sort remittance data, but a person still decides what to send, what to appeal and when to call a payer. Treat AI as a way to make one biller handle more claims accurately.

Which is better for a very small practice?

Outsourcing often suits a practice with one or two providers and no billing staff, because a single absence stops your cash flow. In-house suits a practice with steady volume, a trained biller and a wish to see every claim. Run the cost-per-claim sum with your own numbers before choosing.

What should a billing contract say about leaving?

Notice period, who owns the data, how long you get a full export after exit, who works the open accounts receivable, and whether fees continue on claims you sent before you left. Ask a qualified adviser to read it, and never sign a long term without a performance-based exit.

How do I compare a percentage quote with a per-claim quote?

Convert both to a cost per claim using your own volume. Multiply your average collection per claim by the percentage, add any fixed fees divided by claims, then compare with the per-claim quote plus extras. Confirm whether the percentage applies to gross charges or net collections.

Further reads

• AI Scribe Costs Compared: What Small Clinics Pay in 2026 — What small clinics pay for AI scribes, for the in-house stack.
• Do Small Medical Practices Need an AI Consultant? — Whether a small practice needs outside help with AI.
• Is an AI Scribe Worth It for a Small Private Clinic? — A payback test for AI note-taking in a private clinic.
• Best AI for Dental Insurance Verification: Which Tools Write Back — Eligibility tools that write benefits back to the chart.
• Annual or Monthly Billing for AI Tools: Which Saves You More? — Annual versus monthly billing for the tools you buy.
• Use the AI Already in Your Practice Software Before Buying More — Use the AI already inside your practice software first.
• AI Tools for Physiotherapy Clinics: What Each One Actually Does — What each AI tool does in a clinic setting.
• AI Claim Scrubber: What It Catches and What It Misses — Modifier, bundling and diagnosis edits are caught; eligibility, documentation and fresh payer rules often are not. Test one on 50 claims.
• AI Denial Management for Medical Billing: Group, Appeal, Track — Export your remits, group denials by reason code, draft appeals with AI and track outcomes on one sheet, with a worked month in numbers.
• AI in Healthcare Billing: Patient Statements and Payment Plan Wording — Draft statement and payment plan wording with merge fields, reconciliation checks and a privacy-safe review route.
• ChatGPT for Medical Billing: Appeals and Claim Checks Without PHI — A redaction method, appeal and claim-check prompts, and the review step that lets a billing team use ChatGPT without pasting patient data.
• Medical Billing Proposal: Draft Practice Quotes With AI — Draft a medical billing proposal with AI: percentage vs flat-fee pricing, a scope table, an onboarding timeline and a fully worked quote.
• Using AI for Medical Billing: Spot Missed Charges From Visit Notes — A weekly reconciliation that finds delivered-but-unbilled services, with the spreadsheet step, the AI step and the owner's review.
• AI for Medical Billing Companies: Seven Jobs to Automate First — Seven jobs for a medical billing company to automate first, ranked by risk, with sample prompts, a worked denial-letter example and the human checks each needs.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: fee guides published by OmniMD and Helpware (both sell billing or outsourcing services); CareCloud home and revenue cycle management pages; Steady Medical Billing home page (checked 4 November 2026). Ranges are directional, not audited benchmarks.