Accounts per Biller: The Number to Measure Before You Spend on AI

Every AI pitch to a billing company promises more output. Very few tell you which number to watch. Here is the one we suggest: accounts per biller. It connects directly to how you get paid, it is easy to explain to a client or a buyer, and you can baseline it in two weeks.

Why this number

Most billing companies are paid a share of collections. One pricing guide puts typical fees at 4-10% of net collections overall and 5-8% for most small and mid-sized practices, with specialties like cardiology, orthopedics and oncology at 8-12% (Neolytix, Sep 2026, vendor figure). The same guide lists other models: per-claim fees of $3-$12, flat monthly fees of $500-$2,500, and hybrids such as a base fee plus 3-5%.

Under a percentage model, your revenue grows with your clients' collections, while your main cost grows with the billers you need to work them. So the lever you control is how many accounts, and how many dollars of collections, each biller can carry well.

A simple illustration

At a 6.5% fee, a firm earns about $65,000 for every $1 million it collects for clients. That is arithmetic, not a benchmark. Plug in your own fee and your own collections per biller, and you have the revenue each biller supports.

One public data point shows how thin this can get. A billing-company owner on Searchfunder described a firm with about $1.9 million in revenue, about 240,000 claims a year and about $30 million collected, mostly at a 6.5% fee, that was "barely breaking even" (Searchfunder, post from about 2024). That is one older anecdote from one firm. Do not treat it as typical. But it shows why a small change in biller capacity matters.

The price pressure

An AI-first billing company, Quill Bills, advertises a flat 1.8% fee for its next 10 clinics, against agency fees it puts at "2% to 10%" of revenue (Y Combinator profile, accessed Oct 2026, vendor figure). A competitor pricing that low presumably plans to carry far more accounts per person. You do not have to match their price, but you do need to know your own number.

How to baseline in two weeks

1. Define an account. Usually one client practice, or one provider group within a practice. Pick a definition and keep it.
2. List active accounts per biller. Include partial ownership (a biller who covers half of a large practice counts as half).
3. Time a sample of denials. For each biller, time 10-20 denials from open to done. Do not estimate; time them.
4. Count denials worked per biller per day. From your work queue or billing system.
5. Count claim-status touches per day, and how many needed a phone call.
6. Log hold and call time. Even a rough tally for two weeks shows where hours go.
7. Note aged items. Denials that passed their appeal window, and claims that went untouched for more than a set number of days you choose.

Write it all in one sheet. This is the "before" you will compare against.

How to weigh the cost

On the cost side, list:

• The one-time build: whatever you pay to set up the first queue.
• Tool seats on a HIPAA-eligible plan. Self-serve plans like ChatGPT Business do not come with a business associate agreement (BAA), so budget for a covered plan (IntuitionLabs, Aug 2026, third-party summary). See the BAA checklist.
• Any voice vendor. Published anchors in one vendor article include $10 per hour of call time or $1,500 per provider per month, and a reported $75,000 annual minimum quote from another vendor (Agentman, Aug 2026, vendor article).
• Monthly care: someone has to keep the queue current as payers and rules change.

On the value side, use one plain measure: biller-equivalents of capacity freed. If your baseline says denial work takes a known share of each biller's week, and the queue cuts minutes per denial by a measured amount, you can translate that into hours, and hours into a share of one biller's time.

Then the payback question is simple: one-time cost divided by the monthly value of that freed capacity. Use your numbers, not a vendor's.

A decision checklist

Start a build when all of these are true:

• Your baseline shows denial or status work is a large share of biller time.
• Your systems can export denials, remittances and open claims daily.
• You have past winning appeals to learn from.
• You have, or will buy, a plan covered by a BAA.
• One client practice or payer group is a fair test, and its owner is willing.
• A biller lead will test every step and own the queue afterward.

Wait if:

• You cannot get clean exports yet. Fix that first; it is cheaper.
• You plan to change billing systems within six months.
• Nobody on your team has time to test during the build.

What not to count

• Promised headcount cuts. Whether you hire fewer people as you grow is your decision. Count capacity, not layoffs.
• Vendor figures as your figures. Vendor case studies rarely publish their method. Your baseline is the only number that counts.
• First-month appeal overturns. Appeals take weeks to resolve. Judge the first month on time per denial and denials worked per day.

What to tell clients

When a client forwards a 1.8% offer, a percentage alone loses. Bring your numbers: collection rate, days to payment, denials overturned, and how many accounts each of your billers now carries well. Accounts per biller is the number that lets you defend your price, or lower it on your terms.

If you want help setting the baseline, Lunan's free conversation starts there. We help you pick what to measure and which queue to build first, and you keep the sheet.