AI for medical practices pays off fastest in operations, not diagnosis: answering phones, booking and reminding patients, collecting intake forms, preparing prior authorization requests, drafting visit notes, and checking codes and claims. Every tool that touches patient data needs a business associate agreement, only the data the task requires, and a licensed person checking the output.
We read every regulation, survey and agency page cited here on October 1, 2026. This is general information, not legal or medical advice. Dental offices have their own guide to AI automation for dental practices; this post covers physician practices and clinics.
What can AI do for a doctor’s office today?
AI for healthcare covers two kinds of work. Clinical AI reads images, flags risk and suggests diagnoses. Administrative AI handles the work that fills a practice’s day, and that is where AI tools for doctors are easiest to adopt safely, because a person can check every output before it reaches a patient, a payer or the chart.
| Task | What AI does | Who checks it |
|---|---|---|
| Phones and front desk | Answers approved questions, captures or books appointments, routes clinical calls to staff | Front desk staff review transcripts |
| Scheduling and reminders | Offers open times by visit type and sends reminders | Staff confirm referrals and authorizations |
| Patient intake | Sends forms before the visit and reads returned forms into structured fields | Staff verify identity and coverage |
| Prior authorization | Gathers chart documentation against the payer’s requirements, drafts the request and tracks status | A clinician or authorization specialist submits |
| Visit documentation | An ambient scribe drafts the note from the visit conversation | The clinician edits and signs |
| Coding and billing | Suggests codes from the signed note, checks claims against payer rules, drafts appeals | A certified coder or biller approves |
| Patient messages | Sorts portal messages by type and drafts replies to routine requests | Clinical questions go to a clinician |
How can AI help the front desk and phones?
Phones make a good first project because missed calls are easy to count. An AI receptionist answers the practice line, gives approved answers on hours, locations, accepted plans and new-patient paperwork, and captures appointment requests or books them by visit type where the practice-management system supports a tested connection. Anything about symptoms, test results or medications goes to staff under wording your clinicians approve, and callers hear at the start that they are speaking with an AI.
Our AI receptionist for medical offices typically costs $150 to $300 a month, depending on call volume, hours covered and whether it books directly. Outbound reminder and recall calls need care: in February 2024 the FCC ruled (FCC 24-17) that AI-generated voices count as artificial voices under the Telephone Consumer Protection Act, so those calls need the patient’s prior express consent unless an exemption applies. Our AI appointment scheduling service covers reminders, and AI receptionist vs answering service compares the options.
Can AI handle prior authorization?
It can do most of the preparation, which is where the hours go. In the American Medical Association’s 2025 prior authorization survey of 1,000 physicians, fielded in December 2025, practices reported completing an average of 40 prior authorizations per physician each week, physicians and their staff spent 13 hours a week on them, and 40% of physicians had staff who work only on prior authorization.
AI document automation can pull the diagnosis, history and prior treatments from the chart, check them against the payer’s documentation requirements, draft the request and track it to a decision. The payer side is changing too. Under the CMS Interoperability and Prior Authorization final rule (CMS-0057-F), Medicare Advantage organizations and Medicaid and CHIP programs and plans must, since January 2026, decide expedited requests within 72 hours and standard ones within seven calendar days and give a specific reason for every denial, and they must offer a Prior Authorization API from January 1, 2027. Drugs are excluded. A specific denial reason is what an AI step needs to draft a focused appeal for a person to send.
What does an AI medical scribe do, and is it safe?
An ambient AI scribe listens to the visit through a phone or computer in the exam room and drafts a structured note for the clinician to review, edit and sign. For this kind of AI for medical documentation, the safety questions are practical ones:
- A business associate agreement. The vendor receives the recording and the note, so it is a business associate under 45 CFR 160.103 and needs a BAA before the first visit is recorded.
- Recording consent. State law sets the rule, and it varies. California’s Penal Code §632 requires the consent of every party to record a confidential conversation, whether it happens in person or by phone.
- Retention. Agree in writing how long the vendor keeps audio and transcripts, and who can open them.
- The signature. The clinician who signs the note owns it. Spot-check drafts against the visit, especially medication names and doses.
Where does AI help with medical billing and coding?
AI for medical billing works on three steps. Before submission, it suggests codes from the signed note and checks each claim against the payer’s edits, so a missing modifier or a mismatched diagnosis code is caught before it becomes a denial. After submission, it sorts remittances and denials into work queues by reason. On appeals, it drafts a letter from the denial reason and the chart for a biller to finish. A certified coder or biller should approve every code and claim before it goes out.
If an outside billing company does this work, it is already your business associate, and any AI vendor it sends your claims to is its subcontractor, which needs its own agreement with the billing company. AI workflow automation connects these steps to the systems you already run.
What are the HIPAA rules for AI in a medical practice?
HIPAA has no separate rule for AI. Five existing rules decide whether an AI tool can touch patient data:
- Sign a business associate agreement first. A vendor that creates, receives, maintains or transmits PHI for the practice is a business associate (45 CFR 160.103), and the practice needs a signed agreement before sharing PHI (45 CFR 164.502(e)). Our guide to business associate agreements lists what one must say.
- Keep PHI out of consumer tools. Personal ChatGPT and Claude accounts are not covered by a BAA, and several vendors offer one only on specific business plans. Our comparison of which AI tools sign a BAA lists them.
- Send the minimum necessary. The Privacy Rule requires reasonable efforts to limit PHI to the minimum necessary for the purpose (45 CFR 164.502(b)). HHS guidance says a practice’s policies must identify who needs access to PHI for their job and which categories they need. Write the same thing down for each AI tool and each connection it uses.
- Turn on audit logs and read them. The Security Rule requires mechanisms that record and examine activity in systems that contain ePHI (45 CFR 164.312(b)) and procedures to review that activity regularly (45 CFR 164.308(a)(1)(ii)(D)). An AI tool that cannot show who used it, on which records, does not belong near the chart.
- Update the risk analysis. HHS’s risk analysis guidance says potential risk should be analyzed when a practice plans to incorporate new technology. Add each AI tool, its data flow and its vendor before go-live.
Our AI governance work turns these rules into a short list of approved tools, each with an owner.
What about AI for medical diagnosis?
Diagnostic AI is a different category with a different regulator. The FDA keeps a public list of AI-enabled medical devices authorized for marketing in the United States, each of which has met the FDA’s premarket requirements. The FDA notes the list is not comprehensive, so if a vendor says its tool detects, diagnoses or triages disease, ask for its FDA authorization. Practices covered by Section 1557 of the Affordable Care Act also have a duty under 45 CFR 92.210 to make reasonable efforts to identify patient care decision support tools that use race, color, national origin, sex, age or disability as inputs, and to mitigate the risk of discrimination from their use.
How should a practice start with AI?
- Pick one workflow you can measure. Missed calls, the prior authorization backlog or notes finished after hours each give you a number to track before and after.
- Map the data path. List every vendor the patient data passes through, from the phone carrier or microphone to transcript storage, and confirm a BAA with each one.
- Write the boundaries. Decide what the tool may finish on its own and what always goes to a person, and have a clinician approve the wording for urgent cases.
- Test with fictional patients. Run difficult cases, such as a caller describing chest pain or a denial with missing documentation, before any real patient data goes in.
- Pilot, review, then widen. Start with after-hours calls or one provider’s notes, review the output weekly, and expand only when the numbers hold.
Where NetSys fits
NetSys is a managed IT and cybersecurity company with an in-house AI practice, so for managed clients the same engineers run the network, identities and logs under each AI tool. Our medical office AI receptionist, scheduling and document automation services cover the front office and billing paperwork, HIPAA compliance services keep the BAA register and risk analysis current, and healthcare IT consulting covers decisions such as EHR selection. See our healthcare IT page and the seven-location medical practice whose next compliance review returned 0 findings on its infrastructure portion.
Frequently asked questions
Can staff use ChatGPT for medical questions?
Not with patient information on a personal account. Personal ChatGPT plans are not eligible services under OpenAI’s BAA, so names, dates and other identifiers must stay out. A practice that wants an AI assistant near PHI needs a business plan covered by a signed BAA, and a clinician still checks every answer before it reaches a patient or the chart.
What is the best AI for doctors?
The best AI for doctors is the tool that fits one job and comes with a BAA. For notes, that is an ambient scribe that writes into your EHR; for phones, an AI receptionist with clinical boundaries; for letters and summaries, an assistant on a BAA-covered business plan, such as Microsoft Copilot under Microsoft’s HIPAA BAA. Compare tools on accuracy in your specialty, EHR integration and data retention.
Do patients have to agree to an AI scribe?
Recording consent is set by state law, and some states require every party’s consent. California’s Penal Code §632 does, for in-person and phone conversations alike. Telling every patient before the scribe starts, and noting the answer in the chart, covers the strictest rule. HIPAA’s part is the vendor: it needs a business associate agreement before it receives any recording.
How much does AI cost for a medical practice?
A NetSys AI receptionist for a medical office typically costs $150 to $300 a month, set by call volume, hours covered, providers and locations, and whether it books directly. Scribe, coding and prior authorization tools are priced by their vendors, and NetSys quotes setup work after scoping, with support running month to month.
Sources (checked October 1, 2026)
- 2025 AMA prior authorization physician survey (PDF), American Medical Association.
- CMS Interoperability and Prior Authorization final rule (CMS-0057-F) fact sheet, Centers for Medicare & Medicaid Services.
- FCC 24-17, declaratory ruling on AI-generated voices under the TCPA (PDF), released February 8, 2024.
- 45 CFR 160.103 and 45 CFR 164.502, business associates and the minimum necessary standard, eCFR.
- 45 CFR 164.308 and 45 CFR 164.312, Security Rule activity review and audit controls, eCFR.
- Minimum Necessary Requirement and Guidance on Risk Analysis, HHS.
- List of Artificial Intelligence-Enabled Medical Devices, FDA.
- 45 CFR 92.210, patient care decision support tools, eCFR.
- California Penal Code §632, consent to record confidential communications.
Discuss ai receptionist for medical offices for your business.
Tell us about your current systems, the result you need and your timeline. We will discuss the work, responsibilities and pricing before you decide on an engagement.