ChatGPT for Healthcare: What Your Clinical and Admin Teams Need to Know

I’ve been asked about ChatGPT in healthcare more than any other sector this year. The questions fall into two buckets: what it can do, and what it absolutely should not touch. Let me separate the signal from the noise.

Clinical Teams: Where ChatGPT Fits (and Where It Doesn’t)

ChatGPT can draft patient education materials. It can summarize research abstracts. It can help clinicians prepare presentation slides or translate medical jargon into plain language for families. These are documentation and communication tasks, not diagnostic ones.

Here’s the line: ChatGPT should never touch protected health information. Period. No patient names, no chart data, no case details. The public version of ChatGPT is not HIPAA compliant. Typing “45-year-old male with chest pain and family history of CAD” into a chat window is a violation waiting to happen. Clinical decision support belongs to tools built for healthcare, not general-purpose chatbots.

I’ve seen organizations get excited about AI’s potential and skip the guardrails. The best implementations I’ve seen start with clear boundaries: ChatGPT is for drafting and ideation, not for anything that touches a patient record or informs a clinical decision without human review.

Admin Teams: The Faster, Safer Use Cases

Administrative workflows are where ChatGPT shines in healthcare settings. Scheduling communications, internal policy drafts, employee onboarding materials, meeting summaries, and vendor correspondence all benefit from AI assistance. These tasks don’t involve PHI, and they free up time for higher-value work.

Your HR team can use ChatGPT to draft job descriptions or benefits FAQ documents. Your operations team can prototype process documentation. Your communications team can generate multiple versions of an internal memo in minutes. The ROI is immediate because the risk is low.

The key difference: admin use cases don’t require the same compliance scrutiny as clinical ones. That doesn’t mean zero oversight. You still need guidelines around tone, accuracy, and brand voice. But the regulatory exposure is dramatically lower, which means you can move faster and learn what works without waiting for enterprise AI contracts to clear legal review.

The Conversations Clinicians Should Not Be Having

Let me be specific about what crosses the line. No differential diagnosis brainstorming with patient details. No medication interaction checks using real patient data. No “Can you review this discharge summary?” prompts that include names or dates. No copying and pasting from an EHR into a ChatGPT keynote-style public tool.

The risk isn’t just regulatory. It’s reputational. A single screenshot of a clinician using ChatGPT with identifiable patient information can end careers and trigger investigations. The convenience of a fast answer is not worth the exposure.

If your clinical teams want AI-powered decision support, point them toward HIPAA-compliant tools designed for healthcare workflows. Those exist. ChatGPT is not one of them. The distinction matters, and leadership needs to make it crystal clear before someone makes a costly mistake.

What This Means for Your Organization

Start with policy, not pilots. Define what’s permissible before your teams start experimenting. Separate clinical and administrative use cases in your guidelines. Make HIPAA compliance the non-negotiable line, and communicate it in every training session.

Train your teams on what belongs in ChatGPT and what doesn’t. Give them examples. Role-play scenarios. Make it obvious where the boundaries are. The AI tools keynote I deliver for organizations often focuses on this exact gap: the space between possibility and permission.

Don’t wait for a breach to get serious about AI governance. The organizations that move confidently with AI are the ones that set clear rules early, enforce them consistently, and empower teams to use the technology within safe boundaries.

Joel Comm is a New York Times bestselling author, internet pioneer, and AI keynote speaker. He has spoken for Microsoft, IBM, Cisco, the United Nations, and Harvard Business School. Learn more at joelcomm.com.