The Healthcare Compliance Step-By-Step Podcast · EPICompliance

#109 - AI & HIPAA Compliance: What’s Allowed, What’s Risky, and How to Do It Right

·54 min·3 clips

Jose says “completely HIPAA compliant” is a red flag because one exception breaks the blanket claim.

1. The Healthcare Compliance Step-By-Step Podcast episode focuses on AI and HIPAA compliance, with Ray Walters and Jose Delgado Jr. comparing practical uses and legal risk. 2. Ray Walters hosts, and Jose Delgado Jr. brings 20-plus years in healthcare consulting plus experience running a mental telehealth clinic, which he uses to frame the compliance examples. 3. The episode asks what AI is actually allowed to do in healthcare, what becomes risky when PHI is involved, and how to use it without creating a BAA problem. 4. Jose starts with transcription tools like Dragon and EHR transcription features as the easiest example of AI replacing work already being done. 5. He says transcription is one thing if it only captures spoken words, but it becomes more concerning if the AI coaches what to bill or what to add to the note. 6. He links that concern to claim scrubbing, coding, and the broader difference between an AI that assists and an AI that influences clinical or billing judgment. 7. Jose also points to HR use cases, including writing up sexual harassment and workplace violence issues, where AI can help draft objective language and check whether required points were addressed. 8. He says he likes AI for “completely objective and analytical features” when there is a clear yes-or-no process. 9. When the topic turns to nuance, Jose compares current AI to Wikipedia in 2001 and says nuanced use is still “garbage in garbage out.” 10. He then names patient advocacy, patient education, and translation services as areas where AI can help patients understand medications and conditions. 11. Jose says AI can help patients understand why they are on drugs like metformin and remember to tell providers about other medications. 12. He also raises a concern that a staff member could type patient details into a chatbot and effectively “build the note within the EHR without the EHR.” 13. Jose warns that using AI to generate a clinical impression from partial patient details could look like the tool is “practic[ing] medicine without practicing medicine.” 14. The BAA discussion turns on a basic question: what is the AI creating, storing, or transmitting that makes a contract necessary in the first place. 15. Jose says a clearinghouse can have the BAA, but you are not going to get a BAA with ChatGPT just because it is a popular example. 16. He says the same compliance logic applies if an AI is folded into the EHR, a white-labeled patient portal, or a separate plug-in that touches PHI. 17. The episode’s tone is practical and conversational, with Ray asking follow-up questions from the audience perspective and Jose answering with examples from clinics, portals, and vendor workflows. 18. Jose uses repeated analogies, including Wikipedia, the Terminator, and “garbage in, garbage out,” to make the risk categories easy to remember. 19. Healthcare compliance teams, practice managers, and EHR administrators will get the most from this episode. 20. Viewers wanting deep machine-learning theory or purely technical AI architecture may skip it.

As heard by us

A practical look at where AI fits in healthcare, and where it should not.

Ray Walters and Jose Delgado Jr. keep the conversation practical, rooted in healthcare consulting and mental telehealth rather than big promises.

Read the full review in PlayNext →

Why you'd press play

Before AI talks to your patients, hear where the compliance line gets ugly.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Healthcare Compliance Step-By-Step Podcast