After a 12-hour shift and 24 patients, including two traumas, an emergency physician still has two hours of documentation before she heads home. The notes she writes will determine whether her hospital gets reimbursed for the care she delivered. Each patient carries a different insurance plan, each with its own coverage rules and coding standards. Across hundreds of commercial insurers, thousands of these rules change constantly. If her documentation doesn’t satisfy a payer’s requirements, the claim gets kicked back to a coding team, then a documentation specialist, sometimes back to the physician herself, weeks after the patient left, asking her to clarify a diagnosis she may barely remember. A rule set no physician could memorize, enforced through a process that stretches months past discharge.
Roughly 10-20% of all claim dollars are denied on first submission. The final denial rate, after appeals, is below 2%. That means about 80%-90% of initial denials are eventually overturned. The care was right. The paperwork just had to be fought for, sometimes for months, by teams of people whose entire job is to argue with insurance companies over money the hospital already earned. This is not an indictment of insurers. The system of checks and balances exists for legitimate reasons. It has simply become unmanageable by adding more people to the equation.
What happens between the operating room and the payment
After treatment, clinical notes are translated into standardized codes that insurers recognize. Those codes become a claim submitted electronically. The insurer reviews it against coverage rules that vary by plan, employer, and procedure. If anything is missing or miscoded, the claim is denied. The hospital figures out why, gathers documentation, and resubmits. The process takes 60 to 90 days.
The patient rarely receives one bill; rather, they could receive several bills for the variety of services provided. The emergency physician group bills separately from the hospital. The radiologist bills separately. The anesthesiologist bills separately. A single ER visit generates three or four invoices from different entities, weeks apart, with no explanation of what the patient owes. Every step runs on different software, shares no intelligence with any other step, and costs American hospitals more than 40 cents of every dollar they spend.
$1.4 billion in AI, and the same problems
Healthcare organizations are adopting AI twice as fast as the rest of the U.S. economy and spent $1.4 billion on it last year alone, according to Menlo Ventures. And yet the economics of getting paid have not changed.
The pattern is familiar. HR ran on fragmented tools until Workday connected the full process. IT operations ran on disconnected dashboards until ServiceNow built a platform. Sales teams managed pipelines across three systems until Salesforce unified the workflow. In every case, the winning company built an operating system that connected all steps under a single intelligent layer, and the platform became more valuable with each additional customer. Hospital revenue is the next version of that story.
R1, which manages $80 billion in net patient revenue and serves 95 of the top 100 U.S. health systems, now has the platform to replace the fragmented model.
The architecture behind it
An operating system for hospital revenue management has to connect every step – from the moment a patient is brought through intake to the moment the hospital gets reimbursed, and well after the patient has been discharged. That layer has to understand clinical records and insurance rules simultaneously, learn from every transaction across every hospital it serves, and carry a clear, auditable trail so that clinicians, administrators, and regulators can see exactly why a claim was approved, denied, or routed to a human.
In October 2025, R1 acquired Phare Health, a company founded by former Google DeepMind and Google Health employees. Phare, combined with R1’s R37 AI lab, now powers a system that connects to more than 1,500 insurers and every major electronic health record. It runs on dozens of large language models, architected so any model can be swapped as the technology evolves. And it takes autonomous action across the full payment process: reading clinical records, assigning codes, submitting claims, detecting likely denials before they happen, and generating appeals when they do. People oversee the process to ensure the AI’s reasoning stays accurate, fair, ethical and aligned with expectations.
“At its core, this is a translation problem between clinical language and financial language. That means more than extracting billing codes from a physician’s notes. It requires understanding the full patient record, aligning with documentation, and navigating thousands of insurer-specific rules that change constantly.”
— Dr. Martin Seneviratne, Co-CEO, R1 R37 AI Lab
What makes R1’s position structurally different is the operational data underneath the AI. R1 annually processes 670 million patient encounters and 270 million payer transactions. Each customer’s dataset is used to fine-tune and continually enhance a system for that health system to predict denials before submission, route claims according to payer-specific rules, and learn from every interaction and decision. The system already autonomously resolves roughly 40% of denied dollars, with AI-generated appeals outperforming manual processes while continuously maintaining a human in the loop to monitor and improve the AI delivery. Every decision carries a full chain of reasoning visible to clinicians and compliance teams.
What changes when hospitals get paid in days instead of months
When payment takes months, hospitals sit on billions in unpaid claims. Surgeons spend more time documenting for insurance companies than talking to patients. Patients receive surprise bills weeks after a procedure they assumed was covered.
“Instead of adding another point solution, R1’s R37 Lab is working on AI that can reason across payer policy, coding accuracy, and denial prevention in one connected framework. Real-time adjudication is the goal and the holy grail in this space, and we’re within reach.”
— Lee Kupferman, Co-CEO, R1 R37 AI Lab
When hospitals get paid faster, physicians finish their shifts and go home instead of staying to chart. Patients understand what they owe before they leave the building. Administrators spend their time improving care instead of fighting for reimbursement.
Why Khosla bet on the incumbent
In May 2025, Khosla Ventures, the firm that made the first investment in OpenAI, Block, and DoorDash, invested in R1’s R37 AI lab. A venture firm that built its reputation on category-creating startups looked at a $2+ trillion industry and backed the incumbent.
When the investment was announced, Khosla acknowledged R1’s pioneering use of AI in healthcare, one of the largest sectors of impact for AI, along with its unmatched footprint – serving 95 of the top 100 U.S health systems, with one of the richest and most actionable data environments in healthcare.
The organizations building this architecture now will determine how healthcare works financially for the next generation.
See how R1 is building the operating system for revenue management
