Twelve of the United States’ largest health systems announced in August 2026 that they are forming a Diagnostic AI Consortium alongside Aidoc, a radiology-focused AI company, to jointly build and evaluate how artificial intelligence should be used in day-to-day diagnostic imaging. The members — Advocate Health, Cedars-Sinai Health System, Hartford Healthcare, Houston Methodist, Mercy, Mount Sinai Health System, Northwell Health, Northwestern Medicine, Sutter Health, University of Florida Health, University Hospitals of Cleveland and WellSpan Health — together care for close to 20 million patients annually, making the group one of the largest coordinated AI deployments in American medicine to date.
The problem the consortium is trying to solve
Radiology in the United States is under sustained strain. Imaging volumes have grown steadily for years while the supply of trained radiologists has not kept pace, creating what industry executives describe as a diagnostic capacity crisis — delayed reads, backlogs on urgent scans, and burnout among the clinicians responsible for catching disease early. AI tools that can flag likely findings, triage the most urgent cases to the front of the queue, or draft a first pass of a radiology report have existed for several years, but adoption has been fragmented: one hospital’s AI governance committee may approve a tool that another rejects, and there has been no shared standard for measuring whether these tools actually improve patient outcomes.
What the consortium will actually do
Rather than each health system separately negotiating, testing and validating AI tools, the twelve members will co-design AI-enabled diagnostic workflows together, sharing real-time feedback and measuring impact on safety, quality and speed to diagnosis across all member sites. The goal is to develop shared implementation and governance practices — essentially a common playbook — so that a workflow validated at Northwell Health carries evidentiary weight at Sutter Health or WellSpan Health rather than requiring each institution to reinvent the validation process from scratch. Aidoc is providing the technical backbone through its Clinical AI Reasoning Engine, known as CARE, a foundation model built specifically for diagnostic reasoning, along with aiOS, an enterprise operating system meant to let hospitals manage multiple AI tools under one governance framework rather than a patchwork of point solutions.
Why a consortium instead of a single vendor deal
Aidoc CEO Elad Walach has argued publicly that the urgency behind building a diagnostic AI consortium comes from the scale of the capacity problem: no single hospital system, and no single AI vendor working alone, can generate the volume of real-world evidence needed to prove which AI-assisted workflows genuinely improve diagnosis versus which ones simply add another alert clinicians learn to ignore. By pooling data and protocols across twelve systems with different patient populations, geographies and IT infrastructures, the consortium is betting it can produce evidence robust enough to set an industry standard, rather than twelve separate case studies that competitors can dismiss as anecdotal.
The stakes for patients
For patients, the practical promise is faster answers on urgent findings — a stroke, a pulmonary embolism, a suspicious mass — without sacrificing the accuracy that makes a diagnosis trustworthy. The risk, which the consortium’s governance work is explicitly meant to manage, is the opposite failure mode: AI systems that generate false alarms, entrench bias present in training data, or shift responsibility away from radiologists in ways that erode rather than support clinical judgment. Building shared governance practices before wide deployment is an attempt to get ahead of those risks rather than fix them after tools are already embedded in emergency departments nationwide.
What happens next
The Diagnostic AI Consortium expects to share its initial results in 2027, giving the twelve health systems roughly a year to run coordinated pilots, collect outcome data and refine shared standards before going public with findings. If the collaborative model works, it could become a template for how other high-stakes clinical AI applications — from pathology to emergency triage — get validated across the industry, replacing today’s system of scattered single-hospital pilots with a more coordinated, evidence-driven approach to rolling out AI in medicine.
