The following is an edited Q&A based on remarks from Aspirion’s Liana Hamilton, President, Payment Variance Recovery, and Chase McGrath, Vice President of Product, delivered during a recent webinar hosted by the Healthcare Financial Management Association’s Region 11 chapter. View the webinar recording here.
Q. Let’s start with the big picture. Where does the clinical denials problem stand today?
It’s getting worse, not better. Denial rates continue to climb each year, and the cost and complexity of resolving them at scale is growing right alongside them. Writing a single appeal manually requires expensive, specialized staff—and there’s only so much any one person can produce in a day. For most health systems, that human ceiling becomes the ceiling on revenue recovery.
Q. There’s been a lot of AI hype in healthcare. Are organizations actually seeing returns?
Some are—and the gap between those that are and those that aren’t is widening. Gartner has a term for where a lot of healthcare organizations find themselves right now: the “trough of disillusionment.” Significant investment, uncertain returns, and growing pressure to justify every dollar spent. That’s the reality for many. But the organizations that have built a real strategy—rather than waiting for certainty—are starting to pull ahead.
What we can say is that the results we’re seeing at Aspirion are concrete. Compared to manual appeals, AI-powered appeals are being submitted in 32 days versus 70—that’s 2.2x faster to first appeal. Time to closure drops from 219 days to 158, a 1.4x improvement. And that speed translates directly to cash: clients are seeing time to payment accelerate by 20 days on average.
But speed is only part of the story. On complex clinical denials—the cases most organizations struggle to win at all—we’re seeing a 64% resolution rate. On DRG downgrade denials, which is where we started our AI journey, the resolution rate sits at 27%. Those aren’t vanity metrics. They’re the numbers that show up in recovered revenue. So while plenty of AI initiatives in healthcare are still searching for proof, the performance data we’re seeing suggests that when the deployment is done right, the returns are real.
Q. So what does “being ready” for AI actually look like?
We think about readiness across four pillars: problem-seeking, data foundation, technical infrastructure, and human capital. You need all four. Missing even one is enough to derail an otherwise well-funded initiative.
We also assess readiness across three dimensions—technical, organizational, and regulatory. The goal isn’t to discourage action. It’s to surface the stumbling blocks early, before a solution goes into production. Data quality issues, integration gaps, compliance requirements, change management blind spots—these are all solvable, but only if they’re on the table from the start.
Q. What does implementation actually look like once you’re ready?
The technology alone is never the finish line. What matters is embedding AI into the workflow—routing the right cases to the right expertise, generating first-draft appeals that human reviewers can act on quickly, and tracking submissions against payer deadlines that don’t move. That’s what turns a proof of concept into a sustainable program.
Q. Can you walk through a real example of how this worked at Aspirion?
We started by identifying a clear problem: clinical denial appeals were too complex and too costly to scale with human effort alone. But “automate appeals” was far too broad a goal to act on. So we narrowed it—starting with a small but meaningful segment of our inventory, DRG downgrades for sepsis cases, which represented about 3% of our total account volume.
That wedge worked. Appeal quality was strong, overturn rates were positive, and we had a proof point to build from. From there, we extended the same AI-driven approach across the broader clinical denial universe. That process has now been underway for nearly two years.
The lesson is simple but easy to skip: don’t try to solve the whole problem first. Find a slice that’s meaningful, prove the model, then scale.
Q. How do you measure success before the revenue actually comes in?
That’s a critical question, because payer cycle times can stretch for months. If you’re waiting for financial reports to confirm ROI, you’re losing the internal argument. Leading indicators—resolution rates, days from placement to first submission, overall claim closure timelines—give you a way to demonstrate that the investment is working well before the checks clear.
Q. How do you frame the value of AI to a CFO?
The conventional framing is cost reduction, but I’d push back on that. The more compelling story is revenue enablement: what business can you now pursue that wasn’t economically viable before? For us, AI made it possible to work claims below a dollar threshold that previously made manual processing unprofitable. That’s not a cost saving—that’s a new revenue stream. That’s a much easier conversation.
Q. Build, buy, or partner—what’s the right call?
Each has real tradeoffs. Building gives you control, but it demands significant time, capital, and in-house expertise that most healthcare organizations simply don’t have. Buying is faster, but you sacrifice customization and take on vendor dependency. Partnering—combining RCM domain expertise with outside technical capability—tends to offer the most practical path for organizations navigating lean IT bandwidth and complex compliance requirements.
Q. What’s your closing message for organizations still on the fence?
Payers are not waiting. The volume of denials is accelerating, and the increasingly automated nature of payer-generated response letters is evidence that AI is already deployed on the other side of the table. The question for providers is no longer whether AI belongs in the revenue cycle. It’s whether they have the capital, the capacity, the capability, and the commitment to deploy it effectively. Those that take measured but meaningful risks now are the ones who will see the greatest reward.
To learn more, watch the full on-demand webinar.
