AI in the medical practice: What physicians and administrators need to know in 2026

How widespread is AI adoption among physicians right now?

It is growing fast. According to the American Medical Association's 2024 survey, roughly two-thirds of physicians reported using some form of health care AI in 2024, a 78% jump from the year before. A January 2025 MGMA Stat poll, cited by Physicians Practice, found that AI tools were the No. 1 technology priority for health care organizations, with 32% of respondents ranking them ahead of EHR usability and revenue cycle management systems. That said, adoption is far from universal. A Physicians Practice analysis of the HealthLink Dimensions survey noted that 41% of physicians still do not use any AI tools, and most who do view AI as an administrative assistant rather than a clinical decision-maker.

Where are practices seeing the biggest payoff from AI today?

Documentation is the clear front-runner. Ambient AI scribes, tools that passively listen to patient-provider conversations and draft clinical notes, are the category generating the most measurable results. A multicenter study published in JAMA Network Open in October 2025, summarized by the AMA, found that self-reported burnout among physicians using an ambient scribe dropped from about 52% to 39% after just 30 days. A Physicians Practice feature on ambient listening noted that the technology can cut clinician documentation time by more than 30%.

Beyond documentation, a 2025 athenahealth survey of 501 physicians and administrators found that 60% of AI users said the tools helped them look up clinical information, while 55% said AI consolidated lab and imaging results into a single view. Those findings appear in the same Physicians Practice roundup of practical AI applications.

Are there risks to using ambient AI scribes?

Yes, and they deserve attention. A recent Physicians Practice column cautioned that ambient tools sometimes insert symptoms or diagnoses that don't match the actual patient presentation, or introduce errors such as swapping a patient's sex mid-note. The author's advice: "set it, but check it." Clinicians should review every AI-generated note before signing off, and practices should implement backend validation tools that cross-reference documentation against vetted clinical data. The technology is improving quickly, but human oversight remains essential.

Can AI help with prior authorizations?

This is one of AI's most promising, and most politically charged, use cases. A Physicians Practice survey of its readers found that 58% of respondents named dealing with payers, especially prior authorization requirements, as their top obstacle in 2025. The AMA has reported that physicians now handle an average of 39 authorization requests a week, consuming nearly two full business days of staff time.

On the provider side, AI-powered "agents" can read a patient's chart, pull the required clinical documentation and submit an electronic prior authorization with minimal human input. Early adopters have reported first-pass approval rates above 90% and thousands of reclaimed staff hours per month. On the payer side, however, AI has drawn scrutiny: a 2025 AMA survey found that 61% of physicians believe insurers' use of unregulated AI algorithms is increasing denials and worsening patient harm. A Physicians Practice article on administrative waste in the revenue cycle framed the broader challenge: about 30% of total health care spending goes to administrative overhead, and AI-driven workflow automation is one of the few levers practices can pull to claw some of that back.

What about AI for coding and claim denials?

AI-driven revenue cycle platforms can check coding accuracy, suggest correct codes in real time and flag claims likely to be denied before they are submitted. The financial stakes are significant. As Physicians Practice has reported, roughly $262 billion in claims are initially denied each year out of about $3 trillion submitted, and physicians and staff can spend 10 or more hours a week working those denials. Practices that layer AI into their billing workflows are seeing faster claim cycles and fewer rejections, though published adoption data for small, independent offices remains limited.

My practice is small. Can we realistically afford AI?

Cost is a real barrier. In a Physicians Practice panel on the future of independent practice, family physician Melissa Lucarelli, MD, said her three-provider practice decided to hold off on integrated ambient AI transcription because the price tag, roughly $500 per provider per month for a tool that connects to the EHR, was too steep. Her hope is that AI capabilities will eventually be folded into EHR platforms as a standard feature rather than a costly add-on. That trend is already under way: a Physicians Practice report on new EHRs launched in 2025 described three platforms that are betting on embedded AI assistants, conversational interfaces and tighter integration between clinical and billing workflows.

In the meantime, practices on a tight budget can start with lower-cost or freemium tools, such as general-purpose large language models for drafting patient education materials or summarizing literature, and scale up as pricing comes down.

How should we think about AI and cybersecurity?

Every new AI tool that touches patient data expands your attack surface. A Physicians Practice guide to cybersecurity emphasized mapping how patient data flows through your systems, enabling automatic updates, enforcing multifactor authentication and encrypting all devices. Those basics apply doubly to AI vendors. In January 2025, the HHS Office for Civil Rights proposed the first major update to the HIPAA Security Rule in 20 years, driven in part by the rise of AI. The proposed rule would require covered entities to include AI tools in their risk analyses and mandate vulnerability scanning at least every six months.

As a Physicians Practice feature on breach response put it, cybersecurity is not just an IT problem. It is a leadership responsibility. When vetting any AI vendor, administrators should insist on a signed business associate agreement, confirm the vendor will not use patient data to train its models without authorization and require short breach-notification timelines. A Physicians Practice article on building a scalable cybersecurity program recommended integrating cybersecurity into business strategy from the top down, including regular phishing simulations and tabletop exercises that simulate ransomware scenarios.

Will AI replace physicians?

The short answer is no. A Physicians Practice essay by Neil Baum, M.D., a clinical urology professor at Tulane University, argued that AI cannot replicate the compassion, empathy and communication skills that define the physician-patient relationship. The AMA's survey data bears this out: physicians overwhelmingly prefer AI as a back-office helper, not a clinical substitute. More than half of physicians said reducing administrative burdens through automation was AI's biggest area of opportunity, compared with much smaller numbers who pointed to diagnostics or treatment planning.

That framing was echoed in a Physicians Practice piece on AI's role in empowering independent practices, which argued that AI is best understood not as a replacement for physicians but as a tool for reclaiming time spent on paperwork so providers can focus on patient care.

What's the first step if we want to bring AI into our practice?

Start with a clear problem, not a shiny product. A Physicians Practice guide to implementing AI in five steps advised practices to identify a specific pain point, such as documentation overload, high denial rates or long phone hold times, before evaluating vendors. From there, pilot the tool with a small group of providers, measure outcomes against a baseline and expand only after you have evidence the tool actually works in your workflow. Training matters, too: the AMA found that adequate education and a clear feedback loop were among the critical factors physicians cited for building trust in AI.

A Physicians Practice feature on preventive-care AI made a broader case for viewing AI adoption as a behavioral shift, not just a technology upgrade. The tools work best, that article argued, when they are woven into existing clinical and patient-engagement routines rather than bolted on as afterthoughts.

Where is all of this heading?

Expect AI to become less visible and more embedded. EHR vendors are building AI-powered documentation, coding suggestions and patient-messaging features directly into their platforms. A Physicians Practice case study on AI-driven scheduling described how one health system in rural Georgia used machine learning to predict patient no-shows and automatically create backup appointment slots, an example of AI working behind the scenes without requiring any change in how the physician interacts with the schedule.

The regulatory picture is evolving alongside the technology. Federal agencies are tightening oversight of AI in both clinical and administrative settings, and states are beginning to legislate how insurers can use algorithms in coverage decisions. For practice leaders, the takeaway is to stay informed, invest in cybersecurity and data governance and choose AI tools that solve real problems rather than chasing headlines. The practices that get this right will be better positioned to compete, retain staff and deliver the kind of care that drew their physicians to medicine in the first place.