10 things every practice administrator should know about AI in medical practice

Artificial intelligence (AI) is no longer a horizon issue for U.S. physicians. It’s embedded in electronic health record (EHR) platforms, prior authorization workflows and patient communication tools that practices use every day, often without much fanfare or formal training.

For physicians trying to separate the useful from the overhyped, the landscape can feel overwhelming. These 10 points can help physicians navigate AI on their own terms.

1: You are probably already using AI, whether you know it or not.

Most major EHR vendors, including Epic, Oracle Health and athenahealth, have integrated AI-powered features into their platforms in recent years, from predictive tools to automated prior authorization flagging. Before evaluating any new AI product, take stock of what your current systems already offer. Many practices have AI features sitting dormant that they have never activated.

2: Ambient AI scribes are the technology with the most immediate ROI, but they’re not all the same.

Ambient documentation tools that listen to patient-physician conversations and automatically generate clinical notes have advanced faster than almost any other AI application in medicine. Products from companies including Microsoft-owned Nuance, Abridge, Ambience Healthcare and Suki vary significantly in how they handle specialty-specific language, EHR integration, turnaround time and pricing. Many vendors offer pilots or evaluation periods. Use them before committing.

3: Read every note your AI scribe generates. Every one.

Ambient scribes reduce documentation burden, but they don’t eliminate physician responsibility for what ends up in the chart. Errors in AI-generated notes, such as misattributed symptoms, wrong medication doses or missed negations, carry the same legal and clinical weight as errors in notes you write yourself. Spot-checking is not enough. Develop a fast, consistent review habit before signing off.

4: Understand how your malpractice carrier handles AI.

Most major carriers have not excluded AI-related claims outright. The Doctors Company, for example, has stated publicly that it has no exclusion for AI and would defend and potentially indemnify physicians if AI played a role in a claim. That said, policy language across the industry is still evolving, and carriers are still working through how to assess AI-related risk and underwriting. Before deploying any AI tool in your practice, contact your insurer and confirm specifically how AI-related scenarios would be handled under your current coverage.

5: AI diagnostic support tools are aids, not diagnosticians.

Tools that flag potential diagnoses, surface drug interactions or identify patients at risk for deterioration can add genuine value, but they are trained on data sets that may not reflect your patient population. Overreliance is a documented risk. These tools work best when physicians treat them as a second set of eyes, not a second opinion.

6: Know who owns your patient data when you sign an AI vendor contract.

Many AI tools learn and improve by training on deidentified patient data. Vendor contracts vary widely in how they define deidentification, who retains data rights and whether patient data can be used to train models that serve other health systems or clients. Have legal counsel review any AI vendor agreement before signing, with specific attention to data use and retention clauses.

7: Patient attitudes toward AI are more complicated than they appear.

Research suggests patient openness to AI in health care varies considerably by context and is not universally positive. A 2025 study published in JAMA Network Open found that patients actually preferred AI-generated portal message responses on quality measures but reported lower satisfaction once they knew AI was involved. Separately, a national survey found U.S. adults generally hold low expectations for benefits from AI in health care, with trust in their physician as a stronger driver of comfort than AI itself. What the research consistently supports is that transparency matters: Most patients want to know when AI is being used in their care and how. Make disclosure part of your practice workflow, not an afterthought.

Take our survey: Tell us how AI is changing your practice

We want to hear directly from physicians and practice administrators. Physicians Practice and Medical Economics are surveying readers on their experience with AI scribes and the broader question: Is AI making medicine better?

Your responses will inform our ongoing coverage and may be featured in an upcoming report. The survey takes approximately five minutes to complete.

Please take a moment to anonymously share your experience in the window below, or click here to answer. The story continues below.

8: Front desk and billing AI can save you money before you ever see a patient.

Prior authorization automation, insurance eligibility verification and claims scrubbing tools have some of the clearest, most measurable ROI in the AI landscape. If your practice has not evaluated this category, it is a lower-risk entry point than clinical AI and often pays for itself quickly. Several specialty-agnostic platforms now target independent and small-group practices specifically.

9: AI will not fix a broken workflow. It will make it faster.

Practices seeing the most benefit from AI tools tend to be those that have already done the work of standardizing clinical and administrative workflows. AI layered onto a disorganized documentation process, a fragmented EHR setup or a poorly defined triage protocol tends to amplify existing problems rather than solve them. Before piloting any AI tool, map the workflow it is meant to improve.

10: The regulatory environment is still catching up, and physicians have a stake in shaping it.

The Food and Drug Administration, Centers for Medicare & Medicaid Services and Congress are all actively developing frameworks for AI in health care, with new guidance and proposed rules appearing regularly. The American Medical Association has been pushing for national standards around transparency, bias mitigation and the appropriate apportionment of physician liability for AI errors. Staying informed on the regulatory landscape and engaging through your specialty society when comment periods open matters for the long-term trajectory of how AI gets used in medicine.

Have thoughts on how AI is affecting your practice or medicine as a whole? Click here to take our five-minute survey on AI scribes and the future of medicine.