At $150,000 in monthly collections subject to the fee, 3% is $4,500 a month. Four percent is $6,000. Five percent is $7,500. That is $54,000, $72,000, or $90,000 a year.
Before you get angry about the price, define the work. Before you accept it, ask why that work still costs so much in a world with AI.
Read the denominator
A fee on insurance collections is not a fee on all practice revenue. Gross production is not collections. Minimums change the effective percentage. A tiered headline rate may apply only to part of the collections.
One public example, eAssist, lists 3.5% for single-location insurance collections between $40,000 and $100,000 a month, with graduated rates at higher volumes. That is one provider's structure, not proof that all RCM firms charge 3–5% of total revenue.
Use your contract and actual invoice. Do not compare a percentage stripped of its rules.
Some RCM work is difficult
Incomplete records, denials, and payer disputes do not disappear because software can read an EOB. A good billing team may recover money that would otherwise remain unpaid.
People doing difficult work deserve to be paid. The question is whether you are paying skilled people to solve problems or paying a percentage for repeatable data movement.
Separate verification, claims, posting, denial work, and aged A/R follow-up. What is included? What remains with your office? Is patient billing included? Are clearinghouse charges separate?
Lower fees are not the whole answer
Suppose Provider A charges $6,000 a month and leaves $1,000 of internal work. Provider B charges $4,000 and leaves $3,500 of internal work. B has the lower invoice. A has the lower combined cost: $7,000 versus $7,500.
That hypothetical comparison still needs collections, credits, refunds, write-offs, aging, and denial reasons. Adjust for volume, payer mix, and old A/R before attributing a change to the vendor.
The cheapest service can be expensive if nobody owns the unfinished work.
The model should change as the work changes
The argument is not to replace every person. It is to stop designing every workflow around a person moving information between screens.
Automate repeatable work. Give people a defined queue of exceptions. Measure the outcome. That is how Dentistry Automation's AI-driven, human-supported model should be evaluated.
Run a defined pilot on comparable accounts or locations. Agree on scope, fee base, baseline, internal workload, and reporting. Check accuracy and collections alongside cost. An AI label is not a promise that software will fix every payer problem.
Three, four, or five percent may be defensible for a particular service. It is not self-justifying. Start with the cost of insurance verification and payment posting, then review the full RCM workflow.
See the workflow
Dentistry Automation runs verification, payment posting, and claims workflows with AI, backed by people for the exceptions. See how it works.
Sources: eAssist billing pricing; Dentistry Automation RCM materials. Arithmetic is illustrative, not a Dentistry Automation offer.
