At The Fox Group, we work closely with healthcare organizations navigating emerging technologies, regulatory risk, and operational change. What we are seeing in the field aligns with recent national data: physicians are not waiting for formal enterprise strategies to mature before adopting artificial intelligence. Instead, they are integrating AI tools directly into clinical and administrative workflows, raising important questions about how organizations should respond.
Executive Summary – Key Takeaways
- Physician AI adoption now exceeds 80%
- Clinicians are adopting AI ahead of governance
- Documentation and workflow efficiency drive adoption
- Training, oversight, and policies lag usage
- Healthcare organizations must act before risks escalate
Table of contents
- Physician Adoption of AI Is Accelerating Rapidly
- What are some of the Principal Physician Use Cases of AI Tools?
- What do Physicians Expect from AI Tools?
- What AI Applications are Physicians Using Now?
- What Should Physicians and other Clinicians do Now?
- Frequently Asked Questions About Physician AI Adoption
Physician Adoption of AI Is Accelerating Rapidly
There is little doubt that physicians are at the leading edge of Artificial Intelligence adoption in clinical practice settings.
- A recent AMA survey found that over 80% of responding physicians reported utilizing AI in their everyday medical practice.
- Physician adoption of artificial intelligence has more than doubled in three years, with 81% of physicians now using AI in their practices, according to data collected over two study periods.
- 75% say AI will bring advantages to patient care.
- Diagnostic ability and work efficiency are cited as the greatest benefits.
- The average number of AI use cases per physician has climbed to 2.3 from 1.1 over the past three years.
- Physicians are also concerned about patient privacy and the safety of AI tools used in direct patient care.
- 77% of physicians surveyed also reported using AI tools for personal use at least once a week or more.
- This compares to 57% of people in a national survey employing AI tools for personal use.
- AI adoption by physicians is becoming well-entrenched.
The possible benefits of AI’s impact on clinical practice include improvements in diagnostic capabilities, workflow efficiency, and patient safety, though physicians continue to weigh these against concerns about privacy, clinical skills, and ethical responsibilities.
With adoption now widespread, the more important question is how physicians are actually applying these tools in day-to-day clinical practice.
What are some of the Principal Physician Use Cases of AI Tools?
There are at least seven distinct activities where live use of AI is increasing.
1. Medical Research Summarization and Standards of Care
Physicians are usually lifelong learners, always looking for the latest information on which to base decisions about diagnoses and treatment. Using AI tools to ask about current medical research is a natural extension of this effort.
Standards of care also continue to evolve, making it important for clinicians to stay current on treatment approaches. Almost 40% of respondents cited this as their most common use case.
2. Creation of Discharge Instructions, Care Plans, and Progress Notes
This use case was cited by 30% of respondents as a desirable application of AI. Tools that can review patient care notes on exam findings, diagnostic tests, and medical history, and generate discharge instructions or outpatient care plans are particularly attractive.
These capabilities offer clear opportunities to reduce administrative burden while maintaining clinical quality.
3. Clinical Documentation
Clinical documentation, including billing codes, medical charts, and visit notes, is a core workflow component in healthcare practices. Physicians have expressed strong interest in using AI to support these activities, though only 28% currently report doing so.
Accuracy in procedure coding has long been a challenge for the medical community. AI-assisted documentation can help support both clinical decision-making and billing accuracy, making it a high-value but still emerging use case.
4. Generation of Chart Summaries.
Clinicians are increasingly interested in using AI to generate chart summaries as a way to reduce administrative workload. This use case continues to gain traction as documentation demands increase.
In 2025, 28% of physicians reported using AI for this purpose, compared with only 12% in 2024.
5. Generation of Draft Responses to Patient Portal Messages
Patient portal messages have replaced many traditional phone interactions, but they still represent a significant time burden for physicians. AI-generated draft responses offer a way to improve efficiency while maintaining responsiveness.
Almost 20% of physician respondents reported using AI for this purpose.
6. Translation Services
AI-powered translation services are currently used by 18% of physician respondents. These tools can support communication with patients who speak different languages.
Capabilities such as real-time voice translation could further improve both clinician efficiency and patient experience.
7. Assistive Diagnosis
Only 17% of respondents reported using AI for assistive diagnosis. This relatively low adoption rate may reflect ongoing concerns about the safety and reliability of current tools.
Given the potential influence of AI on clinical decision-making, healthcare organizations must carefully evaluate and manage these tools to ensure optimal patient care.
What do Physicians Expect from AI Tools?
Physicians report generally positive attitudes toward AI tools, with strong expectations for improvements in work efficiency, clinical outcomes, and relief from stress and burnout.
At the same time, concerns remain significant. Physicians worry about potential skill loss, particularly among early-career clinicians; 88% indicate some level of concern about overreliance on AI. There are also ongoing concerns related to privacy, tool reliability, and clinical efficacy.
Questions of liability and accountability further complicate adoption. Determining responsibility for AI-assisted errors introduces legal and ethical challenges, underscoring the need for clear governance and accountability frameworks.
Physicians also want more formal training and education. In fact, 92% expressed a desire for additional training in AI tools, particularly when those tools are embedded within enterprise systems such as electronic health records.
The rapid pace of AI adoption is outstripping institutional readiness, creating a growing gap between real-world usage and formal approval, governance, and training structures.
What AI Applications are Physicians Using Now?
AI adoption is widespread but highly fragmented:
A software company specializing in referral applications surveyed physicians on the types of AI applications they are currently using. They identified what they described as an “adoption paradox”: widespread AI use paired with a scattered, non-standardized toolset.
Physicians are building their own AI tool stacks:
Most physicians report using AI frequently, yet few rely on a single platform. Instead, they assemble personalized stacks, often using personal subscriptions rather than employer-provided tools.
Organizational adoption is lagging behind frontline needs:
Hospitals and health systems continue to move cautiously, while frontline clinicians face immediate pressures such as charting backlogs, inbox overload, and limited time with patients.
Both specialized and general-purpose AI tools are in use:
Some of the most frequently used tools are applied to clinical decision-making and documentation support, including AI solutions such as OpenEvidence and Abridge. More general-purpose tools—such as ChatGPT, Grok, and Claude—are also widely used.
Use cases span both clinical and administrative workflows:
Use cases for these tools range from evaluating diagnostic options to composing patient communications and summarizing clinical interactions.
Patient use of AI is also increasing:
Patient use of AI tools is also increasing. Physicians generally support patient use for routine health and medication questions; 70% believe this use is beneficial or has no impact when used appropriately.
Physicians remain divided on the impact to patient relationships:
However, physicians are divided on the broader impact of AI on patient relationships, with 34% expecting harm and 38% expecting improvement.
Adoption is widespread, but governance remains immature:
This survey reinforces that physician adoption is already widespread, but not yet standardized. Clinicians want greater input into how AI tools are selected and implemented, while organizations must address governance, training, privacy, and the management of shadow tools.
What Should Physicians and other Clinicians do Now?
This situation sounds like the dreaded “Ready, Fire, Aim” approach that many organizations, government agencies, and even individuals find themselves using. AI has quickly become a tool you may feel you cannot do without, but you have very limited time to research and find reliable information about it. And anyway, the health care system you work for has already selected something for you.
The good news is that if you are an individual clinician, you can begin to use off-the-shelf tools such as ChatGPT and Claude for some medical-administrative activities on your own.
Practices should draft a written AI acceptable-use policy to define approved tools and their usage.
If you are a cog in the wheel of a large medical group or health system, try to make sure leadership is taking the issue of governance of the AI tools seriously under consideration for adoption.
The first two enterprise-wide AI tools were released in January 2026.
Take a look at what those offerings are like and be prepared to ask questions about how an enterprise-wide AI might be incorporated into your setting.
In other words, try to get the Aim before the Fire!
Frequently Asked Questions About Physician AI Adoption
Yes, physicians can use general-purpose AI tools for certain administrative and educational tasks, but organizations must carefully manage how those tools are used.
Activities such as drafting communications, summarizing research, or creating educational content may present lower risk. However, entering protected health information into unapproved systems can create significant privacy and compliance concerns.
Healthcare organizations should establish clear policies defining approved use cases, data-handling requirements, and oversight responsibilities.
The most significant risks involve patient privacy, inaccurate outputs, overreliance on AI-generated information, and unclear accountability.
Even high-performing AI tools can generate incomplete, outdated, or incorrect information. When AI influences clinical workflows, organizations must ensure that qualified clinicians review and validate outputs before they affect patient care.
Governance, training, and monitoring programs are essential for managing these risks effectively.
A healthcare AI acceptable-use policy defines which AI tools may be used, how they may be used, and what safeguards must be followed.
The policy should address approved applications, HIPAA requirements, data-sharing restrictions, documentation expectations, human review requirements, and reporting procedures for potential issues.
A written policy helps reduce inconsistent practices and provides a foundation for enterprise-wide AI governance.
The responsibility for clinical decisions generally remains with the licensed healthcare professional, even when AI tools are involved.
AI systems are designed to support decision-making rather than replace professional judgment. Clinicians should independently evaluate AI-generated recommendations and verify important information before acting on them.
Organizations should also establish governance frameworks that clearly define accountability, oversight, and escalation procedures.
Shadow AI refers to the use of AI tools that have not been formally reviewed, approved, or governed by the organization.
This often occurs when clinicians adopt tools independently to improve efficiency or reduce administrative burden. While the intent may be positive, unapproved tools can create compliance, privacy, cybersecurity, and operational risks.
The growing prevalence of shadow AI is one reason healthcare organizations are accelerating governance efforts.
Healthcare organizations should focus first on governance, training, risk assessment, and clear implementation standards.
Leaders should identify approved tools, establish acceptable-use policies, evaluate privacy and security implications, and provide education for clinicians and staff. Organizations should also create processes for monitoring outcomes and addressing emerging risks.
The goal is not to slow innovation, but to ensure AI adoption occurs in a safe, compliant, and sustainable manner.