AI for Medical & Dental Practices

How I’d look at AI in a medical or dental practice

An independent practice runs on its EHR and practice management system, plus a ring of tools around it: phones and texting, patient portal, intake forms, a clearinghouse, maybe an outside billing service, and email. The pressure points are usually the same: a front desk that can’t keep up with the phones, clinicians charting at night, and a billing team chasing denials.

AI can help with all three. It also carries more risk here than in most industries, because the data is protected health information and mistakes can reach patient care. So I start with two questions: where is time being lost, and which tools can touch PHI under a proper agreement.

Start with what you already own

• Your EHR and PM system. The major vendors have been adding AI for documentation, inbox management, and revenue cycle. These features already live inside a system covered by your existing agreements.
• Your patient communication platform. Weave, NexHealth, and similar tools have added AI features for texting and calls. If you already use one, check what’s included before adding another vendor.
• Your office suite. Microsoft 365 and Google Workspace can both be covered by a BAA if configured correctly, which gives staff a safe place for everyday drafting and summarizing.

Where AI genuinely helps

• Documentation. Ambient scribes and note assistants can take a real share of the charting burden off clinicians, if they review what they sign.
• Phones and scheduling. Handling routine calls and texts, filling cancellations, and sending reminders, with a clean handoff to a person.
• Inbox and message triage. Sorting and drafting responses to portal messages for a clinician or nurse to approve.
• Billing. Catching claim errors before submission and drafting appeal letters for denials.
• Administrative writing. Policies, patient education materials, and job descriptions, where no PHI is involved.

Where it’s overhyped or risky

• PHI in the wrong tool. The most common real problem is well-meaning staff pasting patient details into a consumer AI tool. A clear policy and an approved alternative fix most of it.
• Anything that sounds clinical to patients. An AI answering service should not triage symptoms or give medical advice. It should route to a person.
• Notes nobody reads. Ambient scribes can include things that weren’t said or miss things that were. The signed note is the clinician’s responsibility.
• Coding automation without review. Aggressive AI coding suggestions can create compliance exposure. Treat them as suggestions for a trained coder.
• Vague security claims. “HIPAA compliant” on a website is not a BAA. Ask for the agreement.

Questions to ask any vendor

1. Will you sign a Business Associate Agreement, and can I see it before we commit?
2. Where is patient data stored, how long is it kept, and is it used to train your models?
3. For scribes: are recordings kept, for how long, and who can access them?
4. Does it integrate with our EHR, or does someone copy and paste?
5. How does a patient or staff member reach a person when the AI can’t help?
6. What’s live on our plan today, and what’s on the roadmap?

Where I’d start

For most practices: a written AI policy, a BAA-covered tool for staff so they stop improvising, and one focused project, usually either the phones or clinical documentation, whichever is causing more pain. Pilot it with a few willing people, measure it honestly against how things work today, and expand only if it earns its place.