Nuance and Abridge Pass 400 Health Systems; UCLA Trial Shows Most AI Scribes Miss ROI

Scale Without Proof

Nuance DAX Copilot and Abridge have each deployed ambient clinical documentation AI to more than 250 health systems, with Nuance reaching as many as 400. Abridge won Best in KLAS for ambient scribes in both 2025 and 2026. But a UCLA randomized controlled trial covering 238 physicians and roughly 24,000 encounters found that among the tools studied, only Nabla significantly reduced time-in-note. Both tools in the trial showed modest—not dramatic—improvements in burnout.

For enterprise buyers, this creates a gap between vendor scale claims and independent evidence of productivity gains. The market has winners by deployment count, but not yet by controlled clinical outcomes across all vendors.

What the UCLA Trial Means for Procurement

The UCLA RCT is the first large-scale controlled study to measure whether ambient AI scribes actually save physicians time. The headline: most tools did not deliver statistically significant reductions in documentation time. Only Nabla crossed that threshold. Burnout scores improved modestly for the tools tested, but not at levels that justify ROI claims based purely on clinician satisfaction.

This matters because CIOs and CFOs are already being asked to approve multi-year, enterprise-wide contracts with per-provider monthly pricing. At the reported scale—Abridge counting UNC Health, Cleveland Clinic, UPMC, Northwell, Mayo Clinic, Duke, and Johns Hopkins among its deployments—buyers can expect total cost of ownership around 1.7 times the license fee when integration, security reviews, and training are included.

Without hard metrics on time saved per encounter, ROI models collapse. The trial suggests buyers should demand published or site-level data on minutes saved per note, not anecdotes about physician preference. If a vendor cannot show time-in-note reduction in a controlled setting, budget committees should treat claims of efficiency gains as unproven.

Nuance vs. Abridge vs. Nabla

Nuance DAX Copilot, backed by Microsoft and integrated with Epic and Microsoft 365, competes on breadth of EHR compatibility and enterprise bundling with Azure OpenAI. It has the largest reported footprint—150 to 400+ health systems—and the advantage of Microsoft's distribution muscle.

Abridge positions itself as vendor-agnostic with deep EHR integrations across multiple platforms. Its back-to-back Best in KLAS wins signal strong provider satisfaction, a powerful signal in RFP processes where peer references dominate decision criteria.

Nabla, the only tool in the UCLA trial to significantly cut note time, now has the strongest published evidence for productivity gains. It enters enterprise evaluations with a concrete advantage: independent proof that it reduces work, not just redistributes it.

The competitive question for buyers is whether scale and ecosystem integration (Nuance), peer validation (Abridge), or controlled trial evidence (Nabla) weighs more heavily in your risk-benefit calculation. Eachvendorcompetes on a different axis.

Diagnostic AI Delivers Faster Triage, Not Faster Treatment

Barts Health NHS Trust deployed Qure.ai's qXR chest X-ray triage AI via the Sectra Amplifier platform and reported a 61.5% improvement in turnaround time for urgent chest X-rays—13 days to 5 days—and a 41.7% improvement for urgent cancer-case reporting, from 12 days to 7 days.

But independent evaluations show the picture is more nuanced. A Nuffield Trust and NIHR evaluation across 66 NHS organizations found AI imaging tools often yielded modest and varied time reductions in practice. The LungIMPACT RCT in Nature Medicine, covering 93,326 X-rays across five NHS Trusts, found AI reduced reporting time from 47 to 34 hours but did not significantly change time to CT, diagnosis, or treatment.

The gap matters. Faster triage helps radiology workflow. It does not necessarily improve patient outcomes or throughput unless downstream processes—scheduling CT scans, coordinating oncology consults—also accelerate. For enterprise buyers, this means AI imaging tools should be evaluated on workflow efficiency, not clinical outcome improvements, unless the vendor can show evidence beyond reporting speed.

What to Watch

Demand site-level metrics in RFPs. Ask vendors for minutes saved per encounter, time-in-note before and after deployment, and throughput changes at comparable institutions. Anecdotes and satisfaction scores are not enough when CFOs want ROI tied to labor cost reduction or visit volume increase.

Include burnout and throughput KPIs in contracts. The UCLA trial shows burnout improvements are modest. If a vendor sells ambient AI as a retention tool, tie payment or renewal to measurable changes in turnover or hours worked per week.

Watch PACS-layer gatekeepers. The Barts Health deployment via Sectra highlights that imaging AI is shifting from standalone apps to marketplace models embedded in PACS platforms. Enterprise radiology buyers should evaluate whether their PACS vendor offers an open integration framework or locks them into a proprietary algorithm suite. Qure.ai competes with Lunit, Aidoc, Zebra Medical (now Nanox), and native AI from GE, Siemens, and Philips. PACS vendor strategy will determine which algorithms you can deploy.

Expect consolidation pressure. With Nuance and Abridge each deployed across hundreds of health systems, network effects and EHR integration depth will make it harder for smaller vendors to compete on enterprise deals. If you are evaluating a niche vendor, assess their ability to survive a market where the top two or three players control enterprise distribution.