Podcast Summary: Advancing Evidence-Based Care in the Age of AI with Dr. Peter Bonis

Podcast: Becker’s Healthcare Podcast
Host: Erica Spicer Mason (A)
Guest: Dr. Peter Bonis (B), Chief Medical Officer, Wolters Kluwer Health; Adjunct Professor, Tufts University
Date: November 10, 2025

Episode Overview

This episode dives deep into the intersection of evidence-based care and artificial intelligence (AI) in healthcare. Dr. Peter Bonis, an expert in both clinical practice and AI applications, discusses the immense opportunities AI presents for rapid access to medical knowledge, but also critically examines the significant risks and governance challenges this technology brings. The conversation balances enthusiasm for innovation with a sobering look at pitfalls such as accuracy trade-offs, workflow integration, and the continued necessity for trusted information.

Key Discussion Points & Insights

Dr. Bonis' Background & Perspective on AI ([00:34]–[01:44])

• Dr. Bonis shares a summary of his career, including building UpToDate, his clinical practice, and AI safety advocacy.
  • “I took the time to enter myself into Google Gemini … and it was rather sobering because it reduced everything I did for health services in the last 30 years to just a few bullet points.” (B, [00:44])
• Finds AI-generated bios both “interesting and humbling.”

Opportunities & Challenges of AI in Information Access ([01:44]–[07:01])

The Inflection Point of Information Retrieval

• AI enables more efficient information gathering for clinicians & patients.
• However, "Sometimes that efficiency comes at the cost of accuracy.” (B, [02:27])

Trade-Offs Between Efficiency and Accuracy

• High-stakes environments like healthcare require meticulous attention to accuracy.
• “If you just think for a moment at a time when you were sitting in your doctor’s office … how do you know that advice was correct?” (B, [02:54])

Impact of Evidence-Based Resources

• 30% of the time, information from evidence-based resources leads clinicians to change their treatment decisions.

Risks with AI-generated Medical Advice

• Hallucinations (false outputs) persist in large language models (LLMs).

• Even specialists can be misled by subtly or blatantly incorrect outputs.

• Reliability is inconsistent – the same search may yield different responses ([04:10]).

• Biases in LLMs may mirror both societal and cognitive biases, impacting diagnostic suggestions based on input order or content.

  “If you type in shortness of breath at the beginning of a prompt, you might get a different diagnostic suggestion than at the end of the prompt.” (B, [05:00])

Examples of AI System Failures

• Recommending unnecessary surgeries.
• Suggesting stopping medications abruptly with withdrawal risks.
• Advising to avoid safe vaccines.
• Failing to check for pregnancy status before recommending drugs that cause fetal harm.
• “The prospect that some of this information … might sort of slip by and actually get implemented is kind of frightening.” (B, [06:23])

Patient Use of AI for Self-Diagnosis

• Many use LLMs for health information due to barriers to care.
• Raises the question: Can we get speed and accessibility and maintain high quality?
  • “The issue is can we have our cake and eat it too … and actually be able to produce that information but make that high quality?” (B, [06:46])

Ensuring Trusted Information: Governance & Frontline Engagement ([07:42]–[09:29])

Necessity of Strong Governance

• “Having proper governance within a health system or any, any place where you’re deploying technology is just critical and it’s not easy to do.” (B, [07:48])
• Effective governance requires expertise in technology, legal, regulatory, financial, and workflow integration.

Gaps in Frontline Awareness

• Only 18% of frontline providers surveyed were aware of any AI governance at their organization.
  • “If they don’t know of the governance, they’re probably not adherent to the policies … but more importantly, their collective voices have to be understood.” (B, [08:30])
• Involving frontline clinicians is crucial for workflow and policy effectiveness.

Immediate Priorities for Healthcare Leaders: Next 6-12 Months ([09:29]–[12:32])

Variability in Care

• Care quality still varies by geography and demographics; AI must help rectify this.
• “You are entering a healthcare system ... you ideally want to make sure that the system will exit you … with the best possible decisions and the best possible care.” (B, [10:45])

Technology as an Enabler, Not a Replacement

• Goal: “Make it really easy to do the right thing” through safe, effective technology tools.
• Real challenge is the integration of technology into workflows so it benefits both patients and clinicians.

Balance Between Innovation and Operational Realities

• Emphasis on “no margin, no mission”: financial sustainability is key when investing in technology.
• “The challenge is finding that happy nexus between workflow tools which are safe and effective and make it easy for patients to get the right care and providers to do their job well … and to take some of the friction out of the system.” (B, [11:36])

Concluding Thoughts: The AI Investment Boom & the Path Forward ([12:47]–[14:38])

• Immense amounts are being spent in AI compute; expectations are sky-high:
  • “The spend on computer now for the frontier models is around $370 billion per year. … You’d have to essentially be creating $2 trillion of new AI revenue by 2030 ... more than the 2024 combined revenues of Apple and Amazon and Microsoft and Meta and Nvidia combined.” (B, [13:21])
• Not a linear road: suggests the possibility of setbacks or pivots as AI matures in healthcare.
• Key message: High-stakes domains like healthcare demand technologies that are “consistently safe and effective and achieve their purpose.”
  • “That’s where real value is created and real human value is created.” (B, [14:19])

Notable Quotes & Memorable Moments

• “Sometimes that efficiency comes at the cost of accuracy.” (Dr. Bonis, [02:27])
• “Even specialists sometimes can get fooled by [AI’s hallucinations].” (Dr. Bonis, [04:10])
• “The prospect that some of this information that a clinician might be looking at might sort of slip by and actually get implemented is kind of frightening.” (Dr. Bonis, [06:23])
• “Only 18% of frontline providers were aware of any governance taking place around AI or other applications.” (Dr. Bonis, [08:20])
• “Make it really easy to do the right thing.” (Dr. Bonis, [11:32])
• “Being fit for purpose and making sure that you’re doing everything possible to deliver something that is consistently safe and effective and achieves its purpose is critically important.” (Dr. Bonis, [14:18])

Important Segment Timestamps

• [00:43] – Dr. Bonis introduces himself and reflects on AI-generated bios
• [02:27] – Describes efficiency vs. accuracy trade-offs with AI
• [04:10] – Risks of hallucinations and bias in large language models
• [05:00] – Example of diagnostic bias in AI responses
• [06:23] – Real-world failures of AI in clinical decision support
• [07:42] – Importance of governance and frontline engagement
• [10:45] – Addressing care variability across regions
• [13:21] – Scale of global AI investment and what's at stake
• [14:18] – Closing message on the imperative for safe, effective AI

Summary & Takeaways

Dr. Bonis provides a comprehensive roadmap for healthcare leaders grappling with the integration of AI into evidence-based care:

• AI brings speed and new capabilities, but accuracy and trust must never be sacrificed, especially in healthcare where the stakes are human lives.
• Robust governance and including frontline voices are non-negotiable to ensure technology fits real-world workflows and maintains safety standards.
• Rapid adoption and innovation must be balanced with diligence and economic sustainability.
• As the industry races forward with massive investments, Dr. Bonis urges patience and vigilance, echoing his core message: “Consistently safe and effective care is where real human value—and AI’s promise—will be realized.”

This episode offers a nuanced, pragmatic outlook—a must-listen for decision-makers at the intersection of clinical care and innovation.