I didn’t go to medical school to think about billing. But running a sleep medicine practice means I spend a surprising amount of time dealing with claim denials, coding errors, prior authorisations, and payment delays. It’s the unglamorous backbone of keeping a clinic operational, and it’s where a shocking amount of money disappears.

The healthcare revenue cycle — from patient registration to final payment collection — is one of the most administratively burdensome processes in any industry. And it’s ripe for automation.

The Scale of the Problem

Consider an overnight polysomnography. Before the study: verify insurance, obtain prior authorisation, confirm financial responsibility. After: code correctly (CPT 95810 for attended PSG, possibly 95811 if titration was included), submit the claim, track it, respond to insurer queries, post payment, manage the patient balance.

Multiply that by hundreds of patients monthly, across multiple insurers, and you’ve got a full-time job with nothing to do with patient care.

The Medical Group Management Association reports that the average medical practice spends about 15-17% of revenue on billing and administration. Denials alone cost the US healthcare system an estimated $262 billion in rework and lost revenue annually. In Australia, the numbers are different but the inefficiency is comparable — managing Medicare claims, private health fund submissions, DVA requirements, and gap payment collections creates significant administrative overhead.

Where AI Fits In

AI tools for revenue cycle management fall into several categories, and some are more mature than others:

Prior Authorisation Automation

This is where I’ve seen the most immediate impact. AI systems can check insurer requirements, populate authorisation forms with patient data from the EHR, and submit requests electronically. What took our admin staff 20-30 minutes per case now takes 5.

Some platforms predict which claims will need authorisation and initiate the process proactively, reducing day-of-service cancellations from missing approvals.

Coding Assistance

Medical coding is both critical and error-prone. An incorrect CPT or ICD-10 code can trigger a denial, delay payment by weeks, and require manual rework. AI-powered coding assistants analyse clinical documentation and suggest appropriate codes, flagging potential mismatches before claims are submitted.

I’m cautious about fully automated coding — the consequences of errors are significant, and the nuances of sleep medicine coding (distinguishing between diagnostic PSG, split-night studies, MSLT, MWT, and various titration protocols) require domain expertise. But AI as a coding support tool, catching obvious errors and suggesting missing modifiers, is genuinely valuable.

Denial Management

This is where the money really is. When a claim gets denied, someone needs to figure out why, fix the issue, and resubmit. Many denials follow predictable patterns — missing documentation, incorrect patient demographics, expired authorisations. AI systems can categorise denials automatically, prioritise them by dollar value, and in some cases auto-correct and resubmit without human intervention.

We’ve been working with their Melbourne team to evaluate denial management tools, and the initial results have been encouraging. Our denial rate has dropped from about 12% to 7%, and the time to resolve remaining denials has shortened significantly.

Patient Payment Prediction

Some platforms use AI to predict which patients are likely to have difficulty paying their out-of-pocket costs, allowing practices to proactively offer payment plans or financial counselling. I have mixed feelings about this — the intent is good (reducing bad debt while helping patients access care), but the ethics of predictive financial modelling in healthcare need careful consideration.

Implementation Reality

I want to be honest about the implementation experience, because vendor marketing tends to gloss over this.

Data integration is hard. Revenue cycle AI needs access to your practice management system, EHR, clearinghouse connections, and payer portals. Getting these systems talking smoothly takes time and expertise.

Staff training matters. Your billing team needs to understand what the AI is doing and when to override it. Informed trust — where staff understand the tool’s strengths and limitations — is the goal.

ROI takes time. We saw meaningful results at about the four-month mark. The first two months were implementation and training. By month four, the financial impact was clear.

Not everything should be automated. Complex appeals and insurer negotiations still benefit from experienced human billing specialists. AI handles the volume; humans handle the complexity.

Practical Advice for Practice Owners

If you’re considering AI for your revenue cycle:

1. Start with your biggest pain point. If denials are killing you, start there. Don’t try to automate everything at once.
2. Demand integration, not workarounds. Any tool requiring manual data exports or copy-pasting between systems will create as many problems as it solves.
3. Track metrics before and after. Days in AR, denial rate, clean claim rate — measure these before implementation so you can prove the ROI.
4. Involve your billing team early. They know where the real bottlenecks are, and their buy-in is essential.

Revenue cycle management isn’t exciting. But it’s the financial engine that keeps your practice running, and AI is making it meaningfully more efficient. For sleep medicine, where testing and treatment involve ongoing equipment management and insurance complexities, those efficiency gains compound over time.