Abridge built its name on one thing: using AI to listen to doctor-patient conversations and turn them into clinical notes automatically. That was already useful enough to land the company in more than 300 health systems. Now the company wants to do much more.

At a keynote event in New York on June 11, Abridge announced what it’s calling a patient-centered clinician intelligence platform. The idea is to bring AI into every stage of a clinical encounter, from preparation before the visit to billing and claims after it, and to connect health systems with payers and life sciences organizations through those same clinical workflows.

The company also announced a collaboration with NVIDIA to build what both companies describe as the first foundation model built specifically for clinical conversations. Northwestern Medicine, one of Chicago’s largest academic health systems, was named as a new enterprise-wide customer.

How does it work?

Abridge organizes its platform around three stages of the clinical encounter:

• Before the visit: The system generates pre-charted notes and patient summaries tailored to the care setting. For hospital rounds, it pulls context from emergency department notes, nursing assessments, labs, and imaging. For outpatient visits, it surfaces a draft history and flags relevant care gaps or chronic conditions before the clinician walks in.
• During the visit: Abridge now goes beyond passive transcription. It can suggest discussion topics based on the patient summary and the ongoing conversation. Clinicians can query clinical evidence without leaving the platform. The speech recognition supports more than 28 languages across specialties.
• After the visit: The platform generates clinical notes, flowsheets, billing codes, patient summaries, and order drafts. Clinicians can edit outputs using plain language before everything flows into the electronic health record. Abridge integrates with Epic, Oracle Health, and athenahealth.

Abridge is also expanding to nurses. The system can capture nurse-patient conversations during inpatient care and turn them into structured EHR documentation, with relevant findings carried forward into the next clinician’s pre-visit summary.

On the payment side, Abridge is working on real-time claims workflows, aiming to align documentation and billing at the moment care is delivered rather than days or weeks later. The company is partnering with AHIMA, the health information management association, to validate the accuracy and auditability of its coding outputs.

For life sciences, Abridge is exploring how clinical context captured during a conversation could help identify patients who may be eligible for clinical trials, including early screening for conditions like Alzheimer’s disease where biomarkers can appear years before diagnosis.

The NVIDIA collaboration centers on a new foundation model built on the Nemotron open model family, trained on NVIDIA Blackwell infrastructure using de-identified clinical data. Unlike generic large language models adapted for healthcare after the fact, this model is designed to learn clinical reasoning from the ground up across all three training stages.

Why does it matter?

Physician burnout is a well-documented crisis, and documentation load is one of the biggest drivers. Studies have consistently shown that clinicians spend more time on administrative tasks than on direct patient care. Ambient AI tools like Abridge address that directly, and the evidence from early adopters is starting to look compelling.

Reid Health reported that after rolling out Abridge for nurses, its nursing vacancy rate dropped from 18% to 8.6% with no contract staff, and incidental overtime fell 70% on the teams using it. The hospital’s chief nursing officer noted that retraining a single nurse costs close to $100,000, which puts the financial case in sharp relief.

The payment angle is less developed but potentially more significant. Billions of dollars in care goes unrecognized every year because clinical complexity gets lost between the visit and the claim. If Abridge can ground billing codes and claims in a verified record of what was actually said and done, it could reduce the friction between providers and payers that currently slows down reimbursement and creates costly disputes.

The foundation model collaboration with NVIDIA matters because most AI in healthcare today is built on general-purpose models fine-tuned for medical use. A model trained from the start on clinical conversations, across specialties and care settings, could produce more accurate and auditable outputs, which matters enormously in a high-stakes environment where errors have real consequences.

The context

Abridge is not the only company in this space. Competitors including Nuance (owned by Microsoft), Suki, and Nabla all offer ambient documentation tools, and Epic has been building its own AI features into its EHR platform. The market is crowded and consolidating fast.

What Abridge is trying to do with this announcement is reframe the competitive conversation. Instead of competing on transcription accuracy alone, the company is positioning itself as infrastructure for the entire clinical and financial workflow, which is a much harder category to displace.

The clinical decision support additions are also notable. Abridge has signed content agreements with the American Diabetes Association, the American Academy of Family Physicians, the American Heart Association, the New England Journal of Medicine, JAMA, and the Journal of Clinical Oncology, among others. Clinicians using the decision support tools will also be able to claim continuing medical education credit. That kind of ecosystem building takes time to replicate.

Abridge says it now supports more than 100 million conversations annually across health systems that collectively serve more than 250 million patients. Those numbers give the company a significant data advantage as it trains the new NVIDIA model, assuming appropriate de-identification and governance processes hold up to scrutiny. That scrutiny, from regulators, privacy advocates, and the health systems themselves, will be one of the more important tests the company faces as it scales.