Skip to main content
Revenue Cycle Management

AI in the Revenue Cycle Isn't Just for Big Health Systems

8 min read
Dr. Hendrik Lai, Managing Partner, Viturtal Consulting
A practical guide for practices, DSOs and smaller healthcare organizations

I recently joined Melinda Taschetta-Millane of Healthcare Innovation for a webinar on AI in revenue cycle management. I'm Dr. Hendrik Lai, Managing Partner at Viturtal Consulting. The audience was largely CIOs, CTOs and executives, so I expected questions about enterprise platforms and large-scale deployments.

That's not what I got. The questions that came in afterward were far more grounded: Where should a small hospital or ambulatory practice actually start? What return should we realistically expect, and how long until we see it? Why do these projects stall? How do we keep up when insurers are using AI to deny our claims? And will this technology eventually replace our billing team altogether?

Those questions point to a real gap. Most of what's written about AI in the revenue cycle is aimed at large health systems with data science teams and seven-figure technology budgets. But the organizations that feel revenue cycle pressure most acutely are often the smaller ones: independent practices, DSOs, rural hospitals and growing group practices with lean teams and little room for error.

The good news is that AI is not just for big systems. Done well, it may matter even more for smaller organizations. Here's how I'd answer each of those questions.

1. Where to start when you're small

The biggest opportunities sit at the front and middle of the revenue cycle, where errors are born. Historically, organizations have put their people on the back end, chasing claims and appealing denials. But most denials are seeded much earlier, at scheduling, registration and documentation.

For a smaller organization, I'd focus on three areas:

  • Front-end checks. Eligibility and benefits verification, prior authorization status and registration accuracy. These tasks are high-volume and rules-based, which makes them ideal for automation. Every error caught at check-in is one nobody has to chase later.
  • Claim scrubbing and denial patterns. Smaller organizations usually have a handful of payers driving most of their denials. AI that flags high-risk claims before submission, and shows which payer denies which code combinations, gives a lean team a clear place to focus.
  • Appeal drafting. Generative AI can pull the clinical record and payer policy into a first-draft appeal in minutes. For a two- or three-person billing team, that's often the difference between appealing a denial and writing it off.

One practical point: before buying a standalone tool, look at what's already built into your practice management system, EHR or clearinghouse. Fewer systems means less integration work, one fewer dashboard for staff to check, and faster adoption.

2. What ROI to expect, and when

Measure ROI on cash, not on headcount reduction. Headcount is the easy number to reach for, but it undercounts the value and creates resistance from the very people you need to make the technology work.

The hard returns come from:

  • Revenue protected: denials avoided, multiplied by your average claim value, counting the share you would never have recovered.
  • Cash accelerated: fewer days in accounts receivable.
  • Cost avoided: less rework, fewer appeals, less overtime.
  • Underpayments recovered: money that was never flagged as a denial at all.

Then there are the returns that don't show up on the first P&L. The most important for smaller organizations is capacity to grow: adding a location, a provider or a service line without adding billing staff. Close behind is retention. Every time a biller or coder leaves, it can take months before a replacement is fully productive, and in a small team, one vacancy is felt immediately.

On timing, front-end automation tends to pay back first, because the volume is high and the rules are clear. Broader initiatives such as denial prediction take longer, because models need tuning to your payer mix and staff need time to trust the output. I recommend judging the full investment over 24 to 36 months, with checkpoints along the way. Integration and change management carry real costs, and they're often larger than the software license.

3. Why AI projects stall

When an AI project fails, it's tempting to blame the technology. In almost every case I've seen, the real cause is leadership and change management. Three patterns come up again and again.

Automating a broken process. AI is an amplifier. It makes a good process better, but it makes a bad process worse, faster and at scale. Vendors demo their platforms on clean data. Real organizations have duplicate patient accounts, outdated fee schedules and payer rules that live only in one long-tenured employee's head. If a process isn't documented, an algorithm can't learn it.

Buying a tool before defining the problem. "The practice down the street has AI" is not a use case. "Cut eligibility-related denials in half" is. Without a clear target and a baseline measurement, you can't prove the investment worked, and the project quietly loses its budget.

Unclear ownership. AI works by routing exceptions to the right person. If nobody owns a denial category, or two people do, the work just sits. When you ask two people to feed a dog, the dog starves.

Smaller organizations have an advantage here. With fewer people and fewer systems, it's far easier to map a process end to end, fix it and assign clear owners than it is in a large health system.

4. Staying ahead of payer AI

Insurers have moved quickly to use AI in claims review and prior authorization. Their rules now change faster than any manual team can track. For a provider relying on people alone, it's like bringing a human to an algorithm fight.

The pressure shows in the numbers. In Experian Health's 2026 State of Claims survey, 40% of providers said 10% or more of their claims are denied, and 89% named reducing denials a priority. Yet industry benchmarks suggest the large majority of denials are avoidable.

The way to respond isn't to add people to work denials on the back end. It's to move upstream:

  1. Prevent rather than recover. Score each claim's denial risk before submission and route the high-risk ones to a person to fix first. An avoided denial costs nothing; an appealed one costs time and cash.
  2. Learn each payer's behavior. Track denials by payer and by root cause. Patterns appear quickly, often one payer denying one specific code combination, and they can be fixed at the source.
  3. Use AI on your side of the appeal. Let it assemble the record and policy into a strong first draft, so fewer legitimate denials get written off.
  4. Watch net collections, not just denial rates. One 2026 analysis found that initial denials fell while final denials rose and cash didn't improve. Net collections tell you whether you're actually protecting revenue.
  5. Check payments against contracts. Underpayment detection finds money that never shows up as a denial.

The policy environment is moving in providers' favor too, with a growing number of states requiring human review before an AI-driven denial.

5. What happens to your team

One attendee asked directly whether AI will eventually take over 100% of medical billing, with no humans needed. My answer is no. The share of work that needs a person will shrink dramatically, but people remain essential in four places:

  • Exceptions: complex or high-risk claims where judgment matters.
  • Payer relationships: contracts, escalations and challenging denial patterns.
  • Accountability: someone must own what gets billed. "The model said so" won't hold up in an audit.
  • Oversight of the AI itself: monitoring accuracy and catching problems when payer rules change.

The role shifts from processor to reviewer. Your most experienced billers and coders become more valuable, not less, because they know where the problems hide and can tell when the system is getting it wrong.

For leaders, the most important job is to say this clearly and early. If staff believe automation means layoffs, they will quietly work around it. Nothing kills adoption faster than fear. Make it clear the goal is capacity, not cuts, and back that up with a visible career path.

Where to begin

If you're still deciding where to start, my advice is simple: be brilliant at the basics.

  1. Pick one painful, measurable problem. A single denial category, or the one payer costing you the most.
  2. Measure it today. That number is your baseline and, later, your proof.
  3. Map and fix the process underneath it. Write down the steps, clean up the data and assign an owner.
  4. Then apply AI to that one problem, prove the result, and expand from there.

AI won't fix your revenue cycle. But a well-run revenue cycle, augmented by AI, will outperform everyone who skipped the fundamentals. And that's just as true for a three-location practice as it is for a large health system.

Frequently asked questions

Is AI in the revenue cycle worth it for a small practice or hospital?

Yes, if you start with a specific, measurable problem. Front-end tasks such as eligibility verification, prior authorization checks and claim scrubbing are high-volume and rules-based, so even small teams see real time savings and fewer denials. Check what your practice management system, EHR or clearinghouse already offers before buying a separate tool.

How long does it take to see ROI from AI in revenue cycle management?

Front-end automation usually shows results first because the volume is high and the rules are clear. Broader initiatives such as denial prediction take longer. Judge the full investment over 24 to 36 months, with checkpoints along the way, and measure ROI on cash protected and collected rather than on headcount reduction.

What is the biggest reason AI projects fail in the revenue cycle?

Applying AI to a broken process. AI amplifies whatever it touches, so messy data and undocumented workflows produce the same errors faster and at scale. Starting without a baseline is a close second, because without one you can't prove the investment worked.

How can providers keep up with insurers that use AI to deny claims?

Move from recovering denials to preventing them. Score each claim's denial risk before submission, track denials by payer and root cause, use AI to draft appeals, monitor net collections rather than denial rates alone, and check payments against contract terms to catch underpayments.

Will AI replace medical billing staff?

No. AI will handle most routine claims, but people remain essential for exceptions, payer relationships, compliance accountability and oversight of the AI itself. The role shifts from processing claims to reviewing and improving the system.

Working With Viturtal Consulting on Revenue Cycle Performance

Viturtal Consulting helps practices, DSOs and growth-focused healthcare organizations strengthen operations and revenue cycle performance. If you're weighing where AI fits in your revenue cycle, contact Dr. Hendrik Lai at hendrik@viturtal.com or visit viturtal.com to schedule a consultation.

Watch the full Healthcare Innovation webinar, "AI in Action: Transforming Revenue Cycle Management": https://app.webinar.net/wp3GNRAga1B

Related articles

For maximum-threshold RCM benchmarks every practice and DSO should check against — read our complete dental RCM benchmark guide.

For the coding layer gap that clean RCM metrics hide — read our guide to the coding layer and revenue integrity.

Sources: