Healthcare organizations lose an estimated $36 billion a year to denied, delayed, or underpaid claims. Most of that money is never fought for. It disappears quietly, three months after the visit, when a biller looks at a vague note and picks the safest code instead of the correct one.
Revenue cycle AI documentation exists because that failure point is not in the billing department. It is in the exam room, in the seven minutes between a physician asking about symptoms and typing a note nobody has time to get right.
Notiro was built on that observation. Most AI scribes stop at the note. Notiro automates the full clinical day, from patient intake before the visit to ICD-10 and CPT coding after it, so the documentation gap never becomes a billing gap.
Where the $36 Billion Actually Leaks
Revenue cycle management has always been framed as a back-office problem. Fix the claims team, fix the denial workflow, fix the payer contract. That framing misses where the damage starts.
A physician finishes a 15-minute visit covering three chronic conditions and a new complaint. The note gets written in six minutes, often after the patient has left the room. Under that time pressure, physicians consistently document the visit at a lower complexity level than it actually warranted. The Centers for Medicare and Medicaid Services maintains more than 70,000 ICD-10 codes and 10,000 CPT codes, and manual selection under pressure is a documented source of coding errors.
This is undercoding, not a training problem. It is a structural one. The physician who correctly diagnosed and treated three conditions is reimbursed as if they had addressed only one. Multiply that across a full patient panel, and the monthly revenue loss becomes the difference between hiring a second medical assistant and going without one.
What AI Documentation in Healthcare Got Half Right
The ambient scribe category solved a real problem first. Freed, Heidi Health, Nabla, and dozens of others proved that AI clinical documentation could listen to a visit and produce a usable SOAP note in seconds. Physicians got hours back. Burnout scores improved. The American Medical Association has repeatedly identified documentation as the top driver of physician burnout, and ambient scribing has addressed it head-on.
But a note is not a claim. Healthcare AI documentation that stops at transcription hands the coding problem right back to the same rushed post-visit workflow it was supposed to fix.
The physician still has to translate a clinical narrative into billing codes, often without coding training, in the 90 seconds between patients.
What Suki AI and DeepScribe Prove About the Market
Two companies have shown where the real value sits. Suki AI, an enterprise ambient intelligence platform backed by $168 million in Series D funding, markets a 9X return on investment in year one, the strongest financial claim in the category. That number does not come from faster notes. It stems from what happens after the note, when documentation quality directly feeds into cleaner claims.
DeepScribe built its reputation the same way. It holds a 98.8 KLAS score, the highest ambient AI rating KLAS Research has issued, largely on the strength of its oncology-specific revenue cycle integration. DeepScribe pairs ambient scribing with ICD-10 and HCC coding, and that pairing is why enterprise health systems, like those partnered with Flatiron Health, treat it as infrastructure rather than a convenience tool.
Both companies confirm the same thesis from opposite ends of the market. AI revenue cycle management only works when documentation and coding happen inside the same system, not as two separate purchases handled by two separate teams.
The Gap Suki AI and DeepScribe Leave Open
Both platforms also share the same limitation: price and access. Suki AI operates solely through enterprise sales, with every call to action routed to a sales call and no self-serve option for independent physicians or five-provider groups. DeepScribe charges $350 to $500 per provider per month, pricing that puts genuine revenue cycle automation out of reach for the solo practice or small group that would benefit from it most.
That leaves a real gap. A family medicine physician in a two-provider practice loses the same percentage of revenue to undercoding as a hospital system, but has no enterprise contract and no dedicated coding team to catch the error. The practice manager watching monthly collections slip cannot justify a $500 seat license to fix it.
Where Revenue Cycle Management AI Tools Need to Start
Notiro built its coding automation for that exact practice. After the ambient scribe generates the clinical note, Notiro auto-suggests ICD-10 diagnosis codes and CPT procedure codes drawn directly from the visit conversation. The physician reviews and accepts them, then syncs the chart to the EHR in one click, before the next patient walks in.
This is what closes the loop that ambient-only tools leave open, and it is the same coding depth DeepScribe charges enterprise rates for. UCSF research found that physicians using AI scribes with coding support earn roughly $3,000 more per year and see about one additional patient per week, a gain that starts the moment undercoding stops happening by default.
A practice manager running a four-physician family medicine group does not need the infrastructure of a hospital system. A solo internal medicine physician managing diabetes, hypertension, and a new GI complaint in one visit does not have time to hand-select codes for three separate problems. Medical billing AI software has to work at the point of care for the practice that cannot afford a $500-per-provider contract, or it does not solve the problem it claims to solve.
The Exam Room Is the Revenue Cycle
Suki AI and DeepScribe proved that enterprise health systems will pay a premium for documentation that protects revenue, not just time. That validation matters. It means the market already understands that AI-powered revenue cycle solutions are worth building around coding accuracy, not transcription speed.
What remains unsettled is who gets to use them. The $36 billion problem was never isolated to hospital systems with dedicated revenue cycle teams. It runs through every exam room where a physician documents a complex visit in six rushed minutes, whether that visit happens inside a 400-hospital network or a two-physician practice on a main street.
The question every practice should be asking is not whether AI documentation in healthcare can fix undercoding. Suki AI and DeepScribe already answered that. The question is whether the fix is priced for the practice that needs it, or only for the one that can already afford to lose the revenue.
Close the Gap Between the Visit and the Claim
Undercoding is quietly draining thousands of dollars a month from practices that have never seen the number. Notiro’s ICD-10 and CPT coding automation surfaces the codes a visit actually supports, without an enterprise contract or a dedicated coding team. See how it works inside a real clinical day at Notiro.
