At the Becker's CEO + CFO Roundtable in Chicago November 2-5, Dr. David Kirk, CMO of Regard, moderated a panel with Dr. Gurmeet Sran, practicing hospitalist at Sutter Health and former chief clinical data science officer at CommonSpirit Health, and Dr. Faisal Chawla, CDI medical director at Cedars-Sinai. The panelists made the case that getting documentation right at the point of care pays off twice, in the form of better patient care, and in stronger quality scores and revenue.

Key takeaways

1. Documentation gaps start with chart overload. Physicians can only review up to 3% of the chart, and what goes unread never reaches the note.
2. Quality and revenue move together. At Cedars-Sinai, the same encounters showed 7.1% higher risk of mortality and 3.1% higher CMI.
3. Clear notes written during the stay are easier to defend. In most denials, the evidence is in the chart all along, just buried.
4. CFOs should vet the vendor as closely as the tool. Ask questions like what happens to a health system's data if it trains a vendor's future models.
5. Physicians still own every diagnosis. Like a pilot on autopilot, a clinician still has to know how to land the plane.

Strong care can still get lost in the note

Dr. Kirk opened with a story from his years as a health system executive, when a struggling cardiovascular service line turned out to have a documentation problem rather than a care problem. Later, on an ICU shift, a recommended diagnosis caught something that could have seriously harmed his patient, and that convinced him the technology could do more than "move numbers."

Both stories trace back to the same constraint. "As an ICU doc, this is what gets me," Kirk said, "when you go home at the end of the day, and you know that 97% of the chart you have not reviewed." Dr. Sran added that the unread chart hides incidental findings, like a pulmonary nodule tucked into a radiology report, that slip past a hospitalist already juggling 15 other diagnoses.

Cedars-Sinai led with quality, and the revenue followed

Dr. Chawla urged leaders to start with the problem they're solving. For physicians, that often means "pajama time," the hours spent documenting at home after a shift. For Cedars-Sinai, it also meant quality, measured through Vizient metrics, CMS star ratings, and U.S. News rankings.

To isolate the effect, Cedars paired 793 Regard encounters from May and June 2026 with matched twins from the same DRG family, facility, and financial class, with similar lengths of stay. With case mix held constant, Regard encounters showed:

• 17.4% more CCs per encounter
• 12.5% more MCCs per encounter
• 13.3% more conditions documented
• 6.0% higher severity of illness
• 7.1% higher risk of mortality
• 3.1% higher CMI

Higher severity and risk of mortality mean patient acuity is documented more completely, which feeds directly into quality indices. Chawla noted that the CMI gain is significant because comparing within DRG families limits how far it can move. For Cedars-Sinai's CFO, he said, it made the program "kind of a no-brainer."

Every recommendation still has to clear the physician, clinical validation, CDI and coding review, and a compliance check before it reaches the bill. From January through August 2026, 52.3% of Regard's net-new diagnoses at Cedars-Sinai made it through (8,203 of 15,674). The rate climbed with adoption, from about 26% in January through April to about 67% in June through August, well above the roughly 36% baseline across Regard clients. That's a stricter measure than bedside acceptance, which Kirk put at about 95%. Queries fell 6%, and denials, the slowest KPI to move, are still under review.

Payors deny what they can't find

Chawla also handles peer-to-peer calls, where physicians argue to payors that a denied patient needed hospital care. He overturns 70 to 75% of those denials. "The story's there, it's just buried in the chart," he said, and unclear documentation is "making it easy on them."

Clinical validation denials cut closer. Payors accept that the patient needed care but challenge specific CCs and MCCs, sometimes up to four on one account. Chawla's best defense is clear documentation written during the stay. Kirk added that Regard can tune diagnosis criteria to a payor's own policy, so health systems can "put their words back to them."

That's why Kirk argues the work belongs at the point of care, where a diagnosis is easier to defend than one added through a retrospective query. Those queries wear on physicians, too. Sran recalled one arriving during his vacation, with a credentialing warning attached.

Healthy skepticism has to be designed in

Sran warned about "clinical surrender," the slow slide from using AI as a check to leaning on it as a crutch. He has seen AI-drafted discharge summaries with incorrect antibiotic information and urged attendings to reread every draft. He also met a pediatric intern who switched off every AI tool for her first six months to gauge what she was learning on her own.

Sran also pointed to design. Unlike large language models, coded clinical logic, which Regard uses to identify disease states, returns the same result every time, and clinicians can ask the chart why a diagnosis was recommended. As Sran put it, "you work at the speed of trust."

Kirk reminds new physicians that automated EKG reads have offered diagnoses for decades, and a clinician who accepts one blindly is still responsible. "Even if you have autopilot on your plane," he said, "you still have to know how to manually land that plane."

Chawla added that a note serves four functions: communication, medicolegal protection, billing, and synthesis. The last is the most underappreciated, because writing the note is where a physician pulls everything into a clear plan for the patient. That's why he sees it as central to how learners develop, even as more of the drafting gets automated.

What CFOs should ask before they buy

Sran's first question for vendors is data ownership: what happens to a health system's data if it trains future models, and whether the system shares in that value. He also warned against pilot sprawl, recalling a keynote joke about health systems running more AI pilots than Southwest has pilots. Leaders should ask whether a vendor can build one platform across rev cycle, quality, and clinical work, and whether its funding and track record earn trust.

What comes next

Looking one to two years out, Chawla urged CFOs to protect revenue lost to under-documentation while keeping the record accurate. Sran pointed to agentic workflows that comb charts for high-risk patients and to early experiments with AI titrating medications. "It'll be a very different discussion," he said, "if we sit here four years from now."

Kirk closed on the shift he sees coming. As margins tighten and workforce pressure grows, he expects more health systems to move documentation work out of retrospective queries and into the point of care.