—
In under three years, ambient AI scribes, the tools that listen to a patient visit and draft clinical notes, have gone from pilot projects to system-wide rollouts across much of U.S. health care. The case for them starts with a burden every clinician knows: physicians spend nearly two hours on EHR and desk work for every hour with patients, according to a widely cited 2016 study. The case to finance teams is just as direct. An analysis of ambient AI scribe ROIÂ covering 412 deployments found that a license costing about $4,200 a year returned roughly $10,500 for each clinician who actually used it.
Clinicians rate these tools higher than almost any other healthcare software they use. KLAS Research’s 2025 ratings put the category at an average of 93.3 out of 100, compared with 80.6 for healthcare software overall, while Abridge earned 94.7 in February in a field that includes Microsoft’s Dragon Copilot, Ambience, Suki, and Nabla. Those scores, however, capture only part of the experience.
A JAMA study of 8,581 clinicians across five academic health systems found that scribe users saved about 16 minutes of documentation per eight hours of patient care, roughly half an extra visit a week, with no meaningful change in after-hours charting. At Mass General Brigham, the share of clinicians reporting burnout fell from 52.6% to 30.7% after 84 days, although only about a fifth of users completed the final survey. Sixteen minutes may look small on paper, but when a patient is describing chest pain, having the note drafted in the background allows the clinician to stay focused on the conversation instead of typing.Â
KLAS Sub-Grades
The KLAS sub-grades tell a similar story. KLAS breaks each score into six categories, and among the four vendors formally ranked in 2025, product was the lowest or tied for lowest in every case, while relationship and value scored higher. Ambience and DeepScribe had the highest category scores, but KLAS did not rank them because they had only 14 and 7 customer organizations, respectively. Microsoft ranked last despite having the largest customer base in the sample, yet all 34 respondents said they would buy it again.
KLAS research director Mac Boyter advises buyers to look beyond the overall ranking and consider two questions KLAS asks every customer: would you buy it again, and is it part of your long-term plans?
Epic Enters the Ambient Scribe Market
That advice matters after Epic released AI Charting in February, a native ambient tool built with Microsoft’s speech technology that drafts notes and queues orders within the chart. Epic holds about 42% of the U.S. hospital market, so for those health systems, the question is whether a separate scribe contract is worth paying for when the EHR now includes one. KLAS cannot answer that yet because Epic had no customer track record when its 2026 interviews were collected.
How often clinicians use these tools has a direct effect on both the time they save and the return a health system gets from them. The JAMA authors found that clinicians who used a scribe in more than half of their visits saved considerably more documentation time than occasional users, and the financial data follows the same pattern. The $10,500 annual return from the deployment analysis applies to clinicians who actively use the tool, yet only 56.8% of licensed seats were still active a year after go-live, falling to 49% in multi-hospital systems. Once the cost was calculated across every seat purchased, the payback period stretched from 4.8 months to 8.5 months.
For any health system, the real test is whether clinicians are still using the tool six months after launch, something a pilot can reveal before a contract is signed. Epic customers can compare AI charting with a preferred third-party vendor in the same specialty over the same period while tracking active users rather than simply counting licensed seats. Contracts should also give health systems a way to reclaim unused licenses as usage changes. KLAS scores can help narrow the options, but the clearest measure of whether the investment is paying off will come from the health system’s own utilization data.
—
