When I previously led AI at a health system, doctors and coding and clinical documentation integrity (CDI) teams frequently disagreed. Physicians felt the record didn’t capture how sick their patients were. CDI and coding teams pointed out that they could only work with what had been documented.
Both were right. An accurate record matters first and foremost for the patient, but it also drives hospital finances, quality measures, and much more. What’s more, over the years, care has become increasingly complex: over half of U.S. adult hospital stays have three or more comorbidities according to AHRQ. That translation is difficult. Evidence can be scattered across days of notes, tests, medications, and treatment decisions. Because of this, a clinically important detail can be present somewhere in the data in the chart without making it into the final claim.
In my quest to understand what the underlying clinical data could tell us, I started building AI that could one day do deep dives of the patient chart, scaled across every single patient. That led to co-founding SmarterDx.
The care happened. Does the receipt reflect it?
Helping teams close that gap is the work behind SmarterPrebill. SmarterPrebill is clinical AI that reviews every inpatient chart for overlooked diagnoses and coding gaps, giving hospital teams the evidence to validate findings and capture appropriate revenue before the bill goes out.
And it recently hit an exciting new milestone: SmarterPrebill now delivers an average of $4.5 million in realized annual net new revenue per 10,000 DRG discharges. That’s new revenue captured for hospitals and health systems that would have otherwise been left behind.
It took us 18 months to reach $1.5 million, another 12 to reach $2.5 million, eight more to reach $3.5 million, and just four more to reach $4.5 million.
I started working with AI over a decade ago, and I certainly expected AI to be powerful when I started SmarterDx. But I didn’t expect our clinical AI models to improve so quickly. I attribute at least some of that to a critical decision when we founded SmarterDx: we decided to augment human expertise, rather than replace it.
Our customers’ clinical documentation improvement specialists and coders as well as our own physicians and clinical experts work alongside our data scientists, machine learning research scientists, and engineers. The people who understand what a chart should show directly shape how our AI reads it.
Their expert judgment is built into our training data: over time, we have created the world’s largest dataset of over 21 million clinically validated patient encounters, with nearly 1 million more analyzed each month, and more than 3 million expert annotations that teach our models to distinguish when clinical evidence supports a diagnosis — and when it doesn’t. That foundation has helped us deliver each impact milestone above.
The same emphasis on expertise carries through to how hospitals use SmarterPrebill. We believe in narrowly focused AI that strengthens human capabilities while keeping human judgment at the center. Our AI surfaces potential documentation gaps and supporting clinical evidence for CDI and coding teams, reducing time spent searching the chart. Their experts decide which findings are supported and what should change.
Margin gives hospitals choices
More than 40% of U.S. rural hospitals operate at a loss, and 417 are vulnerable to closure, according to Chartis’ 2026 analysis. Financial pressure also limits care long before a hospital closes: positions go unfilled, equipment replacements get deferred, and services become harder to sustain. Hospitals making those decisions cannot afford to lose revenue they have already earned.
That is a problem we can do something about. SmarterPrebill helps hospitals capture appropriate payment for care they have already provided. The hospital has already paid for the nursing hours, medications, and treatment. Recovering missed revenue from that work can improve margin without requiring another admission or asking clinicians to see more patients. Bringing findings and supporting evidence directly to CDI and coding teams also reduces the searching required to make an informed decision.
As a physician, I care about what that margin allows a hospital to do. An additional hire, an equipment replacement, and a service operating at a loss all compete for limited funding. More margin gives leaders more room to support their clinicians and fund the care their communities need. They decide where those dollars go. Our contribution is to help make sure incomplete documentation and coding gaps don’t take those choices away.
The return should be as clear as the recommendation
Hospitals can’t budget against an AI company’s promises. If we say we’ve helped create more margin, their teams should be able to trace that result to the cases that produced it. Our dashboard and case-level reporting make that possible: teams can review the finding, the clinical evidence, the change they validated, and the incremental revenue attributed to that change. We also reconcile results with our clients if any recommendations we put forth are ultimately denied by payers.
That accountability is built into how we do business. The $4.5 million average matters because of the individual results behind it. Each hospital should know its own numbers, and finance, CDI, and coding teams should be able to examine them together. That is what makes ROI useful: a clear accounting of what changed, what it earned, and what it cost.
I’ve described our goal as freeing up “more margin for hospitals to invest in their mission.” That requires us to show exactly what we’ve delivered. It’s why we report realized revenue and make the results open to scrutiny, down to the individual patient case. Hospital leaders should have that level of confidence in the return.
That’s what makes the next milestone worth pursuing. Hospitals already know what they would fund if they just had more room in the budget. I want us to help them move more of those decisions forward, with the revenue they already fairly earned.
Michael Gao, MD, is the CEO and co-founder of SmarterDx. He founded SmarterDx in 2020 after recognizing the significant revenue and quality opportunities hospitals were missing while leading AI at NewYork-Presbyterian. Today, he leads SmarterDx’s mission to build frontier clinical AI to help health systems capture the true complexity of patient care. Earlier in his career, Dr. Gao served as an Assistant Professor of Medicine at Weill Cornell and as Medical Director for Transformation at NewYork-Presbyterian. He earned his BS from the University of California, Los Angeles, and his MD from the University of Michigan, before completing his Internal Medicine Residency and the Silverman Fellowship for Healthcare Innovation at NewYork-Presbyterian/Weill Cornell.
