Days in AR is the metric that quietly kills practice cash flow. You can be booked solid, produce great clinical work, and still watch aging buckets creep from 30 to 45 to 60+ days because nobody has time to sit on hold with a payer or chase a patient balance for the fourth time. Most practices we talk to know exactly which claims are stuck. They just do not have the labor to work them.

This is where AI automation earns its keep. Not by "replacing your billing team" (that framing usually ends badly), but by doing the repetitive follow-up work that a specialist would do if you could find and afford one. Here is what that actually looks like in a real practice.

Where days in AR actually come from

Before automating anything, be honest about where the delay lives. In practices we have worked with, the backlog usually sits in four places:

  • Claims submitted cleanly but sitting in payer queues past their normal turnaround window, with no one calling to check status.
  • Denials and rejections that require a specific action (missing modifier, coordination of benefits, attachment) that nobody has time to rework.
  • Patient balances after insurance, where the first statement went out and then nothing happened for 30 days.
  • Secondary claims that never got filed because the primary EOB landed in a queue nobody watches.

An AI agent will not fix a broken fee schedule or a coding problem at the source. But it will absolutely work the follow-up loop faster and more consistently than a rotating front-desk person doing billing on Wednesdays.

The workflows that move the needle

1. Claim status checks that actually happen

Most practices have a rule like "check on unpaid claims after 21 days." In reality, that check happens when someone has a spare afternoon. An automation can pull your aging report every morning, identify claims past their expected turnaround for each payer, and either hit the payer's portal or eligibility API to fetch status, or in some cases place an automated call through an IVR system to get the current claim state.

The output is not "the AI fixed it." The output is a triaged worklist for your biller by 8am: these 14 claims are still processing, these 6 are denied with reason codes, these 3 show as never received and need to be resubmitted. Your biller starts the day working problems instead of hunting for them.

2. Denial triage and first-pass rework

Denials fall into buckets, and most of the buckets are boring. Missing member ID. Wrong subscriber. Needs COB update. Missing tooth number on a dental claim. Missing NPI on a referral. An AI agent can read the ERA or EOB, categorize the denial, and take the obvious next step: pull the correct info from your PMS or EHR, draft a corrected claim, and route it for a human to click submit.

For the denials that need a real judgment call (medical necessity, bundled procedures, appeals), the agent writes up a summary with the relevant clinical notes and past claim history attached so your biller is not starting from scratch. In practice this cuts the time-per-denial from 15 to 20 minutes down to 2 or 3 minutes of review.

3. Patient balance follow-up that does not annoy anyone

Patient AR is where practices leak the most money, because chasing a $180 balance feels rude and takes time. An automation can run a sequence that most offices simply never execute: statement, then a friendly text with a payment link 5 days later, then an email with the itemized statement, then a live phone call from an AI voice agent that can answer basic questions ("what was this charge for," "can I set up a payment plan") and route anything sensitive to a human.

The tone matters more than people think. We tune these to sound like your front desk, not a collections agency. The result is that patients pay earlier, and the ones who cannot pay identify themselves so you can set up a plan instead of writing off the balance six months later.

4. Eligibility and benefits before the visit

The cheapest denial is the one that never happens. Running eligibility on every scheduled patient 48 hours before their appointment, flagging plan changes, terminated coverage, or new deductibles, and updating the patient's expected out-of-pocket in your PMS prevents a huge share of the rework that eats your AR days later. This is unglamorous work and exactly the kind of thing an agent should do overnight, every night.

5. Secondary claim filing on autopilot

When the primary EOB posts, an agent can automatically generate and file the secondary claim with the primary's payment info attached. This one workflow alone tends to shave a week or more off secondary AR because it removes the "waiting for someone to notice" gap.

What this looks like without hiring

The practices getting real results are not adding a billing FTE. They are combining a lean in-house biller (sometimes part-time, sometimes the office manager) with an agent that handles the volume work. A single office manager can oversee the AR of a two- or three-provider practice when the agent is doing the status checks, the first-pass denial rework, the patient statement sequences, and the secondary filing.

Numbers we see when this is set up well:

  • Days in AR dropping from the 40 to 55 range down to the mid-20s within 60 to 90 days.
  • Patient AR over 90 days dropping by half or more, because the follow-up sequence actually runs every time.
  • Denial resolution time cut roughly in half, since the biller is reviewing prepared work instead of investigating from scratch.

None of this requires ripping out your PMS or EHR. The agents sit alongside your existing systems and act on them the way a human employee would: reading portals, updating records, sending messages, making calls.

The practical setup, and what to watch for

A few things matter more than the software choice.

HIPAA and data handling. Any vendor touching PHI needs a signed BAA, and you should understand where data is processed and stored. Ask specifically about whether prompts and outputs are used to train models (the answer should be no for anything touching patient data). Confirm the specifics with your own counsel and compliance officer, not with the vendor's marketing page.

Audit trails. Every action the agent takes (status check, claim resubmission, message to a patient) should be logged in a way your biller and your auditor can read. If you cannot answer "what did the automation do on this claim, and when" in under 30 seconds, the setup is wrong.

Human checkpoints in the right places. Corrected claims should be reviewed before submission until you trust the categories. Appeals should always be human-drafted or at minimum human-approved. Anything involving a patient dispute goes to a person immediately.

Start with one payer and one workflow. The teams that succeed pick their highest-volume payer and the most repetitive workflow (usually claim status checks) and get that working before adding anything else. Trying to automate all AR at once produces a mess nobody trusts.

Frequently Asked Questions

Do we need to change our practice management system to do this?

Almost never. Good automation works with what you have (Dentrix, Open Dental, Eaglesoft, Athena, DrChrono, Kareo, and others) by using the same interfaces a human would use. If a vendor tells you the answer is to migrate systems first, get a second opinion.

Is an AI voice agent calling payers actually reliable?

For structured tasks like claim status via payer IVRs and some rep calls, yes. For nuanced appeals conversations, no. Use voice agents for the high-volume, script-friendly calls and keep humans on the judgment calls. The math still works out heavily in favor of automation because 80% of the calls are the boring kind.

How do patients respond to AI voice or text follow-up on balances?

Better than most practice owners expect, if the tone is right and the agent can actually help (take a payment, explain a charge, offer a plan). Patients get frustrated with dead-end phone trees, not with useful automated conversations. Always give a clear path to a human.

What is a realistic timeline to see AR improve?

First results (usually in patient AR and claim status turnaround) show up in three to four weeks. Full impact on days in AR typically lands at the 60 to 90 day mark, once the backlog of aged claims has been worked through and the ongoing workflow is steady state.

What happens if the agent makes a mistake on a claim?

The same thing that happens when a person does: you catch it in the audit log or on the next EOB, you correct it, and you tune the rule that caused it. This is why human review on corrected submissions matters early on, and why logging every action is non-negotiable.

Where to start

If your days in AR is north of 35 and you have been telling yourself you need to hire a billing specialist, run the numbers on automating the follow-up loop first. In most practices we have worked with, the ROI shows up faster and the operational load on your existing team goes down, not up. If you want a concrete look at what would work in your specific setup (your PMS, your payer mix, your current backlog), talk to our team at Qintara Corp and we will walk through it with you.