AI Scribe Lifts Coding 27% While 70% Diagnoses Lack Transcript

The Money Arrives

Ambience Healthcare closed a $243 million Series C on July 29, 2025, at a $1.25 billion valuation — unicorn territory less than five years after Mike Ng and Nikhil Buduma founded the company in 2020. Oak HC/FT and Andreessen Horowitz co‑led; the OpenAI Startup Fund, Kleiner Perkins, and Optum Ventures re‑upped. New capital came from Frist Cressey Ventures, Town Hall Ventures, Smash Capital, Georgian, and Founders Circle Capital. The round brings total funding to more than $343 million, combining the Series C with the $100 million previously raised from investors including Kleiner Perkins, OpenAI Startup Fund, Andreessen Horowitz, Optum Ventures, Human Capital, Martin Ventures, AIX Ventures, AirTree Ventures, and individual backers such as John Doerr, Jeff Dean, Richard Socher, Pieter Abbeel, Anne Wojcicki, Eren Bali, Jay Desai, Nish Bhat, and Matt Mochary.

Ambience's platform is live at more than 40 health‑care organizations — Cleveland Clinic, UCSF Health, Houston Methodist, Memorial Hermann, St. Luke's Health System, John Muir Health among them — and three‑quarters of clinicians at its largest academic medical center customers use the product daily, covering four‑fifths of visits. Those utilization numbers, which the founders cited in an a16z Bio & Health interview, explain why investors moved from "can technology serve this need" to "how fast can you scale it."

KLAS validated the traction with a 97.7 customer satisfaction score, top marks for product quality and responsiveness. Ambience also claims the first third‑party‑validated, CFO‑approved ROI tied to coding accuracy and compliance — a distinction that matters when health‑system boards demand payback periods measured in quarters, not years.

Inside the Hospital Floor

Most ambient AI scribes were built for outpatient settings. In an outpatient clinic, the pattern is predictable: a clinician enters a room, talks with a patient, and the AI turns that conversation into a note. That model collapses inside a hospital. A hospitalist juggles ten or more patients at once, each at a different stage of an evolving stay. Diagnoses shift. Labs trend. Medications change daily. In roughly half of all inpatient encounters, the patient is sedated, cognitively impaired, or otherwise unable to give a reliable history — so no conversation occurs to capture. The data that actually drives care lives in the chart: labs, imaging, medication changes, prior diagnoses. Ambience's own analysis found that 70 percent of clinically important inpatient diagnoses have no signal in the transcript at all. For the CC/MCC conditions that determine appropriate reimbursement, that percentage rises further.

The company recognized this gap early. Rather than retrofit an outpatient tool, Ambience designed its inpatient suite around hospitalist workflows from the ground up. In February 2026, it expanded chart awareness to key features across the platform. By late May, that awareness was enabled end‑to‑end across the full inpatient workflow — admission through discharge. The result is a system that doesn't just listen; it reads.

"The majority of what happens in a hospital stay is invisible to ambient scribes. We built a system of intelligence that can see and understand the full clinical picture," the company stated in its May 2026 announcement.

The feature set maps to each phase of a hospital stay. Patient Summary runs nightly, ingesting structured and unstructured data from the patient's chart and Epic's CareEverywhere network, up to 18 months of history across all visit types, and distills each encounter into two to three essential bullets. Before the hospitalist enters the room, the context is waiting. H&P Notes generate the full history and physical from the bedside interaction, when one occurs, fused with chart data. The Diagnoses engine surfaces missed conditions and precise ICD‑10 codes with supporting evidence in real time, drawing from the conversation, the hospital problem list, and months of chart history. Progress Notes build each day's note from the prior day's note, the current bedside interaction if it happened, and real‑time chart data — ending the copy‑forward habit that propagates stale findings. Hospital Stay Summary synthesizes the entire admission for clean handoffs between care teams. Discharge Summary constructs the final summary from the complete stay in one click, with the option to enrich it via a recorded discharge conversation.

All of this lives inside Epic. Ambience sits directly in the Epic Toolbox, accessible through Hyperdrive and Haiku with no copy‑paste, no manual upload, no toggling between systems. It uses the Epic Ambient Module and native FHIR APIs to read and write into the chart. Early deployment data at St. Luke's Health System in Idaho, where Patient Summary launched in June 2025, shows utilization across H&P notes, progress notes, and discharge summaries exceeding 70 percent, with Net Promoter Scores rising 31 points. In a separate evaluation across 66 encounters at four health systems, chart‑aware discharge summaries resolved 91 percent of information gaps that standard documentation had missed. Nearly one in five Ambience employees is an active or trained clinician, embedded across product development, model training, and clinical validation — a staffing model that keeps the product grounded in the workflow it serves.

How Cleveland Clinic Chose

Cleveland Clinic's decision to standardize on Ambience followed a six‑month head‑to‑head pilot pitting five major ambient scribes against each other across more than 80 specialties and subspecialties. Over 300 clinicians participated. Each solution was tested for three months. Physicians with varying comfort levels with technology rotated through 20 specialty areas, generating 25,000 encounters on Ambience alone. The clinic measured cognitive load using the NASA Task Load Index and the Mini Z burnout instrument — not vendor‑supplied dashboards. In February 2025, Cleveland Clinic announced an exclusive five‑year partnership.

"We are committed to finding innovative ways to support our providers and reduce administrative burden and burnout while enriching patient experience," said Beri Ridgeway, M.D., Enterprise Chief of Staff. "Our clinicians reported their enjoyment with more face‑to‑face patient interaction, and that the software provided detailed notes that improved care coordination across specialties. Importantly, they also felt less burnout."

The scale of that deployment is substantial. Cleveland Clinic employs 5,743 salaried physicians and researchers across a 6,690‑bed system spanning 23 hospitals and 276 outpatient facilities in Ohio, Florida, Nevada, Toronto, Abu Dhabi, and London. In 2023 the system logged 13.7 million outpatient encounters and 323,000 hospital admissions. Adoption data from the pilot tells its own story: providers opted to use Ambience‑generated documentation in 80 percent of visits — two to three times the rate of competing scribes, per the clinic's comparative analysis. Sixty‑seven percent of clinicians reported reduced cognitive burden. Same‑day chart closure rose 7 percent. Face time with patients increased 32 percent. "Pajama time," after‑hours charting, dropped nearly 50 percent. Note creation time fell 25 percent. Specialty‑level adoption varied: nephrology hit 99 percent; urology lagged at 55 percent. When cardiology adoption stalled at 50 percent, Ambience engineers worked with the clinic to upgrade specialty‑specific features, pushing utilization to 71 percent.

Houston Methodist, a nine‑hospital system, has been piloting Ambience in both the emergency department and inpatient settings with full Epic EHR integration. "Joining Epic's Toolbox program reflects our deep commitment to providing the power of Ambience to clinicians, integrated deeply within their workflows," Buduma said. Epic confirmed the integration in its own Toolbox announcement, expanding clinician access to ambient voice recognition tools across its installed base.

Meese, Cleveland Clinic's executive director of digital health, said she has shared the findings with other systems that reached out after the ViVE 2025 presentation. The clinic also plans post‑mortem reviews with the four vendors it did not select.

The Clinician's Day

Ambience's dashboard shows clinicians spending 45 percent less time charting overall and 45 percent less time on after‑hours documentation. Utilization across deployed sites averages 80 percent. In head‑to‑head bake‑offs against two unnamed competitors, Ambience hit 80 percent adoption while Vendor 1 stalled at 26 percent and Vendor 2 reached 40 percent.

"Ambience has been the most transformative thing we've done at John Muir Health," said Priti Patel, MD, Chief Medical Information Officer at the Northern California system. That assessment carries weight because John Muir deployed the platform across outpatient, emergency, and inpatient settings. The platform now covers more than 200 specialties, including oncology, psychiatry, and emergency medicine, domains where generic scribes have historically failed.

EHR integration amplifies the effect. Ambience lives inside Epic's Toolbox, Hyperdrive, and Haiku, no copy‑paste, no toggling, no manual upload. When the scribe writes the note in the same window the clinician signs it, the 45 percent reduction holds in practice. The vendor comparison data suggests the market is separating. High utilization combined with high time savings indicates clinicians voluntarily keep using the tool, the strongest signal that the product fits the workflow rather than disrupting it. Competing scribes that require heavy editing or specialty‑specific workarounds show lower adoption. The arms race in ambient AI is no longer about who can transcribe; it is about who can disappear into the clinician's day.

Where Coding Meets Revenue

Ambience's pitch has always been that documentation and coding are the same workflow, not adjacent ones. The company's own benchmark makes that concrete: its model was tested against 18 experienced, board‑certified physicians on ICD‑10 coding accuracy and showed a 27 percent relative improvement over the expert baseline.

During an encounter, the platform listens, structures the clinical narrative, and simultaneously suggests ICD‑10 and CPT codes in real time. It surfaces Hierarchical Condition Category (HCC) opportunities that risk‑adjustment models depend on, and it guides Evaluation and Management (E/M) level selection so the documented complexity matches the billed level. All of this happens inside the note while the clinician is still in the room, or, in the inpatient setting, while the hospitalist moves between beds. The company describes it as "revenue integrity starts with documentation that reflects the full clinical picture."

Chart summarization sharpens the effect. The Patient Summary feature, launched at St. Luke's Health System in Idaho, reads the full chart, prior admissions, specialist notes, imaging, labs, and produces a concise pre‑visit briefing. That context feeds the ambient scribe so the encounter note captures diagnoses that might otherwise be missed. The summarization also flags gaps before discharge, giving CDI teams a chance to query while the patient is still on the unit.

For health systems, the financial logic compounds. More accurate ICD‑10 and CPT capture reduces denials and downcoding. Real‑time HCC identification supports risk‑adjusted contracts, Medicare Advantage, ACOs, commercial value‑based arrangements, where missing a single hierarchical condition can shift a patient's risk score and the corresponding capitated payment. E/M guidance prevents both underbilling and overbilling. And because the coding suggestions are embedded at the point of care, the compliance burden shifts from retrospective chart review to prospective confirmation. The clinician accepts, edits, or rejects; the auditor sees a contemporaneous trail.

Early adopters report the cycle time shrinks. Instead of a coder chasing a physician days later for clarification, the note arrives with structured codes already attached. That reduces accounts‑receivable days and the labor cost of the coding department itself. Ambience has not published a system‑level ROI figure in the open literature, but the component metrics, 27 percent coding accuracy lift, real‑time HCC and E/M support, chart‑wide summarization, map directly to the levers that drive net revenue per discharge. The company reports $13K in revenue per clinician per year at St. Luke's, verified by AAPC and KLAS.

The Competitive Board

Nuance's hiring signals reinforce the priority. Its research lab in Seattle lists open roles for founding technical recruiters, RL researchers, model optimization engineers, and ML infrastructure specialists, all oriented toward AI research and engineering rather than product delivery. The open requisitions cluster around core model work, suggesting the Microsoft subsidiary is investing in foundational capabilities that could feed both Epic integration and standalone Dragon Ambient eXperience (DAX) products.

Suki is taking a different tack: deep Epic integration rather than platform ownership. The company's own site claims its Epic integration delivers ambient notes, ICD‑10 coding, and clinical Q&A inside Epic with two‑way sync. Suki's marketing language, "goes far beyond transcription, capturing the entire patient …", mirrors the chart‑aware positioning Ambience has staked out for inpatient workflows. Digital Health News noted that Ambience "joins AI ambient scribe players Suki and Commure in the program," a phrasing that implicitly groups the three as the current tier‑one contenders.

For health systems, the choice increasingly hinges on workflow fit. Epic's native tools will be frictionless for organizations already standardized on Epic; Nuance brings the installed base of Dragon users; Suki offers a modular layer that promises coding and Q&A inside the chart. Ambience's bet, chart‑aware intelligence spanning admission to discharge, validated by a 27 percent relative improvement in ICD‑10 coding accuracy versus board‑certified physicians, targets the inpatient complexity that outpatient‑first scribes still miss.

What the Growth Buys

Capital at Ambience's level buys more than runway; it buys a hiring mandate. Ambience has 15 open roles with a salary band running $42,000 to $350,000 (median $265,000). Two positions were posted in the past week alone: a Staff ML Engineer, Frontier AI at $250,000–$350,000 and a Head of Enterprise Accounts at $250,000–$300,000. The roster also includes Head of National Strategic Accounts ($250,000–$300,000), Staff Engineer, Agent Productivity ($250,000–$300,000), Staff Software Engineer, Distributed Systems ($250,000–$300,000), and Staff Software Engineer, Product ($250,000–$300,000).

Those titles signal a dual push, deeper model work on the research side and hardened enterprise sales muscle on the go‑to‑market side. Every incumbent integration raises the bar for what a standalone platform must deliver, real‑time coding awareness, inpatient chart summarization, pre‑visit prep, and each capability requires specialized talent: clinical NLP researchers, FHIR integration engineers, revenue‑cycle product managers, and security architects who can satisfy health‑system procurement reviews.

The ROI narrative, coding accuracy gains of 27 percent relative to board‑certified physicians, documentation time savings that clinicians say cut burnout, is what justifies the salary bands now appearing on job boards. Health systems are buying ambient AI the way they once bought EHR modules, as infrastructure. The vendors that staff up fastest will set the integration standards everyone else must follow. The next time a hospitalist opens a chart at 6 a.m., the summary waiting there will have been written by the company that hired the engineers to build it.

Working in frontier tech? Zero G Talent tracks the openings: see every open Ambience Healthcare role, browse frontier tech jobs, openings at Nuance Labs, and the people building the field.