How AI is Reshaping Clinical Practice, Governance, and Patient Trust
The Clinical Pathways Congress and Cancer Care Business Exchange (CPC+CBEx) panel “Beyond the Hype Part 1: Real-World Impact of AI-Powered Clinical Documentation,” brought together three distinct voices—health care attorney Adria Warren, JD, partner at Foley & Lardner, LLP; Johnetta Blakely, MD, executive vice president of quality and clinic operations at Tennessee Oncology; and Alti Rahman, MHA, MBA, CSSBB, chief strategy and innovation officer at the American Oncology Network. Each panelist highlighted how artificial intelligence (AI) is being introduced into clinical settings and the broader challenges of adoption, governance, and patient trust. Together, their perspectives underscored both the promise and the complications of bringing AI into the physician’s office.
Adria Warren opened by situating AI within the legal and compliance landscape. While contracts can define guardrails, true impact happens at the point of care where technology meets patients and providers. Legal frameworks, however, are still underdeveloped. Few health systems have dedicated AI governance protocols, and emerging state laws in California, Colorado, Utah, and Illinois are creating a patchwork of rules. Warren emphasized compliance structures such as transparency, auditing, and accountability, echoing federal guidelines. She also warned of liability concerns: ultimately, physicians remain responsible for the accuracy of AI-assisted documentation, and contracts must reflect that reality.
Dr Blakely provided a frontline perspective. Initially skeptical, she described how ambient AI tools like DeepScribe have been transformative, freeing up nearly 2 hours daily and capturing nuances that might otherwise be lost in notes—such as social details or medication counseling. AI has also streamlined workflows by digesting more than 1 million annual faxes into electronic medical records. In pilot testing, automation eliminated thousands of redundant clicks, directly improving efficiency. Yet Blakely cautioned that AI is not for everyone; some physicians remain uncomfortable with outputs that do not “look like their notes.” Importantly, she stressed the need for human oversight. AI can propose diagnoses or codes, but clinicians must decide what is clinically relevant.
Adding a systems-level view, Rahman explained how adoption strategies vary depending on market saturation and resource availability. Rural practices, for example, need scalable tools that solve access challenges, such as reducing wait times for appointments, while competitive urban markets demand more sophisticated solutions. He described the “buy vs build” model for technology: Certain tools such as DeepScribe are purchased, while others such as structured data insights are developed internally. He emphasized the importance of trust, rigorous vendor security assessments, and validation measures like the F1 score to confirm AI accuracy. Continuous auditing, often by trained nurses, is used to ensure outputs remain reliable.
Patient perception emerged as another recurring theme. Blakely reported that most patients readily consent when told that AI tools are being used to document visits, often welcoming the efficiencies. However, issues of privacy, consent, and data ownership remain unresolved. While de-identification offers some protection, panelists noted the risk of re-identification in small datasets. Patients themselves are increasingly experimenting with generative AI tools like ChatGPT, raising new expectations for physicians to respond to AI-informed questions.
Finally, the discussion turned to payer relations and value-based care. Rahman argued that AI could help bridge the gap between clinical documentation and payer decision-making, particularly in reforming flawed prior authorization processes. By surfacing timely, structured evidence, AI might enable “gold carding” arrangements that reduce unnecessary barriers to approved therapies.
In conclusion, the panel presented AI not as a distant future but as an evolving reality with measurable benefits—time savings, improved documentation, and streamlined workflows. Yet success depends on governance, validation, and maintaining the human element in clinical judgment. The consensus was clear: AI is a tool, not a replacement, and its adoption will hinge on building trust among clinicians, patients, and regulators alike.
Reference
Warren A, Blakely J, Rahman A. Beyond the hype part 1: real-world impact of AI-powered clinical documentation. Presented at the Clinical Pathways Congress; September 5, 2025; Boston, MA.
