A few years into AI’s rollout in medical group practices, the productivity question has moved past the pilot stage. Practice leaders who signed contracts and shepherded adoption are now asking sharper questions: Are these tools delivering returns in the operational metrics that sustain a practice — patients seen, documentation completed, revenue captured, and rework avoided — beyond improvements in clinician enthusiasm or satisfaction scores?

What you told us

Our May 12, 2026, MGMA Stat poll put that question directly to medical practice leaders: Have new AI tools made your providers more productive in the past two years? Less than half (46%) say they have. Another 27% report no productivity gains from AI, 14% are unsure, and 13% say they do not use AI at all. The poll had 257 applicable responses.

The topline numbers already suggest a complicated story; open-ended responses show how textured it really is.

• In practices reporting gains, ambient AI scribes dominate as the primary driver of productivity. Respondents repeatedly named tools such as DAX, DeepScribe, Suki, and similar documentation solutions for speeding note completion easing administrative burden. Coding assistance, clinical decision support, and generative AI tools (for denials, summaries, and patient messaging) came up too, but far less often and usually mentioned as complementary tools.
• Where AI hasn’t moved productivity, leaders pointed to adoption and integration friction: resistance to change, inconsistent use, training needs, and poor EHR interoperability that breaks the workflow. Cost and implementation burden came up frequently, along with doubts about the tools’ maturity — particularly around accuracy, specialty fit, and day-to-day practicality. Several leaders in this group told MGMA that even where AI cuts documentation time and burnout, it hasn’t translated into more visits or measurable productivity, which feeds their skepticism.
• The “unsure” group is largely a measurement problem. Many practices are still implementing AI, waiting for access, or lack visibility into who is using the tools and how often. That inconsistency, paired with limited tracking and feedback, makes it hard to establish a productivity baseline, let alone measure change. Some leaders cited anecdotal drops in administrative burden, but the impact on visit volume, RVUs, or other measurable productivity metrics isn’t clear yet.
• Most non-users plan to adopt AI within the next 12 months, drawn by expected gains in efficiency, productivity, cost reduction, and patient satisfaction — and many are already researching or piloting AI scribes and scheduling tools. A smaller group is hesitant or opposed, citing leadership disinterest, clinical or data sensitivity concerns, and dissatisfaction with their existing EHR-native AI features. Others are limited by practice structure (e.g., hospital-based settings) or general uncertainty. Intent to adopt is broadly high; readiness and feasibility vary.

What the research shows and where it gets complicated

The research base on AI and clinician productivity has matured considerably in the past two years, but the findings are more nuanced than many vendor pitch decks suggest. The studies fall into two camps: those measuring what clinicians experience and those measuring what shows up on the balance sheet.

The burnout and well-being evidence is strong

The most consistent finding across recent research is that AI documentation tools — ambient scribes in particular — reduce clinician burnout and improve how they experience their workday. A multicenter quality improvement study published in JAMA Network Open in October 2025, covering 263 physicians and advanced practice providers (APPs) across six health systems, found that ambulatory clinicians using an ambient AI scribe saw burnout drop from 51.9% to 38.8% within 30 days. The same study reported meaningful improvements in cognitive task load, after-hours documentation time, and the ability to stay present with patients during visits.

A stepped-wedge randomized pragmatic trial published in NEJM AI in late 2025 confirmed those results. Across 66 practitioners in ambulatory clinics, ambient AI significantly reduced work exhaustion and interpersonal disengagement — a coprimary outcome measuring the emotional toll of documentation overload. A separate randomized clinical trial in the same journal tested two commercial AI scribe products against usual care across 238 outpatient physicians across 14 specialties over eight weeks. Both tools reduced time-in-note, and physician survey measures of burnout and task load improved.

These are among the first rigorously designed randomized trials of these tools in ambulatory settings.

The productivity and financial evidence is weaker

The financial and throughput evidence tells a more complicated story. A notable NEJM AI editorial published in November 2025 titled its assessment bluntly: “AI scribes are not productivity tools yet.”

Written by researchers at UC San Diego and Kaiser Permanente, the editorial reviewed two of the landmark randomized trials and concluded that AI scribes save roughly seven to 22 minutes per provider per day — with the caveat that some of those estimates may be inflated because they rely on EHR signal data rather than direct observation. The authors warned that health systems expecting scribes to free up enough time for additional patient visits — and thereby pay for themselves through incremental revenue — may be disappointed by what the current evidence supports.

A January 2026 cohort study published in JAMA Network Open by UCSF researchers compared physician revenue, patient volumes, and claim denials between AI scribe adopters and nonadopters within one health system. The study was designed to answer the question most practice leaders care about: did the tool move the financial needle? The results were sobering for those expecting immediate ROI from documentation AI.

A separate longitudinal study from a large Midwestern health system tracked 220 primary care clinicians and more than 314,000 patient encounters over 150 days after AI scribe adoption. The findings showed gradual improvement: documentation time savings grew from about 7% on day one to 15% by day 150, and billed wRVUs edged up by roughly 2%. Critically, these changes weren’t instant — they emerged as clinicians adapted to the new workflow. The payoff from AI adoption is not an on-off switch; it’s an adaptation curve.

A February 2026 study from a large academic medical center, published in the American Journal of Managed Care, added another important nuance. Nearly all ambulatory physicians using an AI scribe perceived that their documentation time had decreased — but actual measured time savings were modest on average and concentrated among providers with the highest documentation burden to start with. The perception of improvement often outpaced the measurement. Keep that in mind as you evaluate vendor claims.

A 2025 rapid review in JMIR AI synthesizing real-world evidence on digital scribes echoed the mixed findings: billing-based productivity metrics were largely unchanged across the studies reviewed, while self-reported documentation time decreased and physician engagement improved.

The Peterson Health Technology Institute Assessment

In March 2025, the Peterson Health Technology Institute released what remains one of the most comprehensive independent assessments of AI scribe technology, drawing on interviews and convenings with health system leaders, AI vendors, and industry experts. Its core conclusion: AI scribes are likely reducing clinician burnout, but the financial impact remains unclear. Provider organizations reported no consistent increase in patient throughput attributable to the technology. Some noted that providers felt they could take on additional patients, but that sense hadn’t translated into measurable volume gains. PHTI also flagged that clinician adoption rates, once a tool is broadly available, tend to plateau between 20% and 50% — well short of universal uptake.

Beyond ambient scribes

Ambient scribes have consumed most of the oxygen in the AI productivity conversation, but other categories are entering ambulatory workflows. AI-assisted coding tools are being deployed across revenue cycle operations, with vendors reporting gains in coding speed and accuracy. AI-driven clinical decision support is advancing through both EHR-embedded tools and standalone platforms — though a January 2026 report from the Stanford-Harvard ARISE network cautioned that AI model performance often degrades when moved from controlled benchmarks into the ambiguity of live clinical encounters.

The Doximity 2026 State of AI in Medicine Report, based on surveys of more than 3,100 physicians, found that literature search and voice-based documentation were the two most common AI use cases. Adoption of ambient tools rose from 20% to 29% of physicians surveyed between early 2025 and early 2026. Three-quarters of physician AI users said the technology had already reduced administrative workload and improved job satisfaction — but again, that’s a self-reported experience metric, not a financial productivity measure.

The AMA's 2026 Physician Survey on Augmented Intelligence found that 81% of physicians now use AI in some professional capacity — more than double the 38% rate in 2023. The most common uses: summarizing medical research, documenting clinical encounters, and generating care plans. Physicians were most optimistic about AI's potential to improve work efficiency and diagnostic accuracy, but they also flagged data privacy (86%) and robust safety and efficacy validation (88%) as prerequisites for broader adoption. Confidence is growing, but trust still has conditions.

The distinction that matters most

The most important distinction in this evidence base is between two kinds of ROI.

The first is experiential/well-being ROI: providers feel less burdened, spend less time on after-hours documentation, report lower cognitive load, and stay more present with patients. This is real and valuable — it affects recruitment, retention, morale, and clinical culture. It is also the outcome that the strongest evidence supports.

The second is financial/productivity ROI: providers see more patients, generate more wRVUs, improve coding accuracy, reduce claim denials, or free enough time to add meaningful clinical capacity. This is often the outcome needed to justify sustained investment — and the one where the evidence is thinnest. Early signals exist, including the modest wRVU gains in the Midwestern longitudinal study and scattered health system reports of improved coding capture. Financial productivity benefits may emerge in time. But the gap between vendor promises and peer-reviewed evidence remains wide.

One emerging concern deserves attention. A policy brief in npj Digital Medicine in late 2025 raised the issue of an "AI coding arms race," warning that ambient scribe vendors are increasingly positioning their tools as revenue-cycle platforms that drive higher-complexity billing codes. This shifts the value proposition from reducing clinician burden to optimizing revenue capture — and raises the questions of whether some of the financial gains attributed to AI are genuine productivity improvements or documentation-intensity changes that could invite payer scrutiny.

How to evaluate AI productivity claims

To help sharpen your evaluation criteria as you consider or renew AI tools, the questions below are designed to help separate credible productivity claims from marketing:

• Ask vendors for pre-post evidence using objective measures, not just satisfaction surveys. The research consistently shows that perceived documentation time savings outrun measured savings. If a vendor cites only survey-based improvements, the productivity claim is incomplete.
• Demand clarity on whether reported wRVU or revenue gains reflect true throughput or coding-intensity changes. A higher average E/M code does not mean a provider is seeing more patients or working more efficiently. It may mean the tool is capturing documentation elements that were previously undercoded — a legitimate benefit, but a different one than productivity.
• Ask what the adoption curve looks like and how long before gains stabilize. The longitudinal data shows that AI scribe benefits evolve over 150 days or more. Numbers from a 30-day pilot may not hold at scale.
• Identify which providers benefit most. The evidence suggests that providers with the highest baseline documentation burden see the largest objective time savings. A blanket productivity claim across all specialties and provider types likely overstates the case for many of your clinicians.
• Separate burnout reduction from productivity. Both matter, but they are not the same business case. If your primary goal is clinician retention and well-being, the evidence supports investment. If your primary goal is additional patient capacity or revenue, the evidence may not be there yet at the same level of confidence.
• Understand the total cost of implementation: workflow redesign, training time, IT support, and the ongoing editing burden on providers. Several studies and health system leaders have noted that AI scribes shift providers from being note authors to note editors — a genuine workflow change, but not necessarily a time elimination.

Conclusion

AI tool adoption is accelerating, and the evidence base is maturing fast. The strongest current case for these tools right now is that they make clinicians’ workdays better — and for many practices struggling with recruitment and retention, that’s reason enough to invest. The case that AI tools are reliably making clinicians more productive in the financial and throughput sense that practice balance sheets require is still being built.

Leaders who go in clear-eyed about that distinction will make better purchasing decisions, set more honest expectations with their clinical teams, and be better positioned to capture real gains as the technology and the evidence continue to evolve.

Join the conversation 

• MGMA Stat polls are conducted weekly to give medical practice leaders a pulse on the latest trends in healthcare management. To participate, sign up for MGMA Stat at mgma.com/mgma-stat. 
• Have a success story in implementing AI to address physician burnout or improve productivity? Let us know in the MGMA Member Community or email us at connection@mgma.com.