The 90-day claims fix: how a 4-op practice cut AR over 90 by 38%

A four-operatory general practice in the upper Midwest signed up for Pulse in February. Their AR over 90 days was $112,400. Ninety days later it was $69,800 — a 38% reduction, no new staff, no carrier renegotiation. This is the un-cherry-picked version of how that happened, including the parts that didn't work.

The starting line

What the books looked like in February.

Annual production: $2.1M. AR total: $284k. AR over 90: $112k. Their billing coordinator — call her Dana — was running point on collections three afternoons a week, posting ERAs by hand on the other two, and working denials whenever the queue caught fire. She is excellent. She is also one person.

The practice ran two reports out of Open Dental — Aging by Patient, and Aging by Carrier. The first was 18 pages long. The second came out clean but nobody had time to mine it for patterns. Both reports told them what was old. Neither told them why, or which 20 accounts mattered most.

What we changed

Three things, in this order.

One — surfaced the right 20. Insights ships a default board called "Top 20 oldest, by collectibility". It's the cross-product of dollar amount, age, and our model's estimate of whether the carrier will pay. Dana stopped working the alphabetical list. She started working the surfaced list. Two of her three afternoons a week went from triaging to actually calling.

Two — split the queue. Insurance AR and patient AR were collapsed into one giant pile. Pulse splits them. Insurance AR moved to Remit's denial queue; patient AR stayed with Dana. The denial queue auto-drafts the appeal narrative — she edits and sends, doesn't draft from scratch. That alone gave her back about four hours a week.

Three — bank-feed the EFTs. Eleven percent of their AR-over-90 was… already paid. The EFT had hit the bank and never been reconciled against a ClaimPayment. Pulse bank-feed picks that up automatically now. The first week we flipped that on, $14k of "outstanding" insurance AR quietly reconciled itself.

What didn't work

Two things we tried that produced exactly zero.

Patient-payment text reminders. We bolted on the SMS reminder integration. It moved the patient-AR number by about $1,200 over 90 days. Not nothing, but the staff time to configure it was greater than the dollar return. Their patients pay at the front desk. Different practice, different answer.

Auto-resubmit denied predeterminations. We had it on for 60 days. Most carriers came back with the same denial, just slower. We turned it off and pointed Dana's manual appeal workflow at the high-value cases instead. Better result, less noise.

The 90-day numbers

AR over 90: $112,400 → $69,800 (-38%). AR total: $284k → $241k (-15%). Collections ratio (collected / produced): 94.1% → 96.7%. The collections ratio is the one that matters for a healthy practice — every point of recovery here is roughly $21k/year at their production rate.

The boring lessons

You almost never need more staff to fix AR-over-90. You need the right 20 accounts surfaced, insurance and patient queues split, and the bank feed actually reconciling. Most practices we audit have at least one of those broken. Many have all three.

The other lesson is that the "AI" part of this story is small. Remit's appeal-drafting saved hours; Insights' surface-the-top-20 saved more. Neither is a model writing original prose. They're an automation system pointing the human at the highest-leverage work, and then doing the rote drafting after.

You don't need a smarter billing coordinator. You need a system that lets the one you have stop triaging.

Practice details lightly anonymized at the customer's request. Numbers are real and verified against the OD ledger across the 90-day window.