Creating an actionable framework for making informed AI investment decisions
Hospitals are facing a new kind of denial—one that’s harder to detect, harder to fight, and quietly draining millions from the bottom line. Payers are no longer just rejecting claims for missing codes or paperwork; they’re using AI to retroactively question the clinical validity of services that were already approved and delivered.
"Payers are using advanced analytics to revisit and deny claims based on their own interpretation of clinical appropriateness long after the care was given. It’s turning every claim into a moving target," says Jim Bohnsack, chief strategy and client officer at Aspirion. "And the volume is increasing in a way that hospitals simply can’t keep up with using people alone."
As payers automate denials at scale, hospitals must respond in kind—deploying AI not as a future strategy, but as an immediate necessity to preserve revenue and protect care decisions.
Here, Bohnsack unpacks how hospitals can use AI to meet payers on equal footing. He also explains what finance leaders should look for when evaluating AI investments. Finally, he shares how Aspirion’s Doc IQ platform is helping providers scale their response, reduce appeal volume, and strengthen financial performance amid increasingly complex denials.
Q: What are the most impactful ways AI is improving revenue cycle performance for hospitals and health systems right now?
Bohnsack: One of the biggest impacts is in addressing the increase in denial volumes. These aren't typical denials where the claim had a form error. Payers are shifting the rule sets, and we’re seeing a rise in denials tied to clinical documentation. They’re publicly saying that they’re pulling back in areas like prior authorization to reduce friction, but what’s really happening is they’re shifting that scrutiny to the backend, resulting in increased requests for medical records. Newly applied technologies and processes are questioning the value and medical necessity of the care provided. Upon receiving medical records, payers ingest the raw data into large language models (LLMs) and apply rule sets to deny or downgrade claims retroactively. It’s the same outcome: they are still reducing their overall claim spend.
That leaves providers with only one option: a credentialed expert, such as a coder or nurse, must comb through the full medical record and build a compelling appeal. Most providers can’t keep up. Payers know this, are better capitalized, and can apply technology more quickly. However, AI is leveling the playing field. At Aspirion, what we’ve always done in a human-centered way is utilize nurses and lawyers to pull the raw medical record, review and compare it against care and coding guidelines, and craft comprehensive appeals. But in 2023, when the volume became too high, we built Doc IQ, a platform powered by LLMs that replicates what our people do best. It pulls the clinical evidence from the hundreds or even thousands of pages contained within a medical record, matches it to payer policies and care and coding guidelines, and generates a complete argument.
As a result, we’ve seen a decrease in the number of appeals required, indicating that the first appeal is so comprehensive that it overturns the denial. Our success rate has also improved dramatically, as has the speed to resolution.
Q: Can you share a specific client example where AI made a measurable difference?
Bohnsack: When we started building our AI platform, we worked with a handful of early partners to identify their biggest pain point: DRG downgrades related to sepsis. For many of our clients, this is a frequent and costly denial type, particularly associated with the argument of whether a person had sepsis or not. We used AI to build appeals that directly addressed disputed DRG downgrades for sepsis, resulting in a significant improvement in the success rate.
From there, we expanded to all DRGs, then to multiple disputed codes, and finally to patient type, such as inpatient versus observation. We kept broadening the coverage on the types of clinical denials and appeals our clients were experiencing. Now, rule sets vary by payer, and we can accommodate the vast majority.
Q: For healthcare finance leaders evaluating AI investments, what metrics should they focus on to determine success?
Bohnsack: First, clearly define the problem you're trying to solve. Is it increasing coverage and throughput volume? Or is it improving success rates and total recovery? Those become your baseline metrics. It’s critical to understand the current state and the limitations of the human-driven process and then assess the cost to develop a solution.
At Aspirion, we looked at all those areas. Our baseline expectations were that the technology had to perform as well as humans in terms of throughput, speed, and success rate, with the end goal being an appeal letter to overturn a denied claim. Once we met that bar, we focused on increasing volume with the same or fewer people, while keeping a high success rate.
We didn’t try to automate everything at once. First, we structured the medical record to make it easier for humans to review and search. Then we taught the system to pre-answer a series of questions the humans always ask when assessing a case, including the reason for admission and who the consulting physicians were, which led to a summary cover sheet that a person can easily digest. Each piece improved efficiency while allowing us to validate quality at every step. You don’t need a fully baked solution to start seeing ROI.
Q4: What’s the typical timeline for seeing ROI, and what kind of improvements in key metrics are your clients experiencing?
Bohnsack: For our clients, ROI has been measurable from the beginning, and it strengthens as the platform scales. Our early focus on DRG downgrades tied to sepsis delivered immediate improvement in overturn rates and validated that this type of clinical denial was the right place to start. That work set the foundation for broader application, and today, clients are seeing results across a much wider range of denial types.
In January, Doc IQ processed about 2,000 of the 20,000 clinical appeals we handled. Last month, it handled more than 14,000, over two-thirds of our clinical volume. The platform now supports roughly 70% of clinical denial types, with a path to more than 90% in the next quarter.
AI performance continues to improve, and the biggest impacts are appearing in efficiency and outcomes. We’ve seen about a 15% decrease in the number of appeals required per case. We’ve also seen a 10-20% increase in success rates. And we’ve reduced the time from placement to payment by approximately 40 days—a change that shows up directly in AR days and cash flow for our provider clients.
These results aren’t theoretical. Clients are seeing fewer appeals required, more denials overturned, and revenue recovered sooner. That kind of measurable impact is very real to a VP of revenue cycle trying to stay ahead of rising denial volumes with limited resources.