AI receptionist for medical offices: what it should handle, what it costs and how to set it up

The short answer

An AI receptionist for a medical office is a voice agent that answers the calls your front desk cannot reach, books or requests appointments in your EHR or practice management system, captures refill, records and referral requests, and transfers anything clinical or urgent to a person. It earns its keep at opening, over lunch and after hours, when hold times climb and callers hang up.

Buy one only if three things are true: the vendor signs a business associate agreement (BAA) that covers the AI, recordings and transcripts; it books into your EHR (or creates requests your staff confirm each morning); and it follows red-flag rules your physicians approve. Packaged products cost roughly $114 to $2,000 a month in published 2026 pricing, while a custom agent on a usage-based voice platform runs on per-minute fees plus a one-time build. The rest of this guide shows how to decide.

Why medical office phones are the bottleneck

Phones are still where a practice loses the most staff time. In an MGMA Stat poll of 10 March 2026 (294 responses), practice leaders named the most time-intensive phone tasks for their staff as eligibility and prior authorization (45%), scheduling (31%), intake (9%) and prescription refills (6%). Most of the scheduling, intake and refill calls follow a script, which is exactly the work an AI receptionist can take over. Eligibility and prior authorization usually need a different fix, covered below.

Staffing makes it harder. The US Bureau of Labor Statistics reports that healthcare and social assistance employs 42% of all US receptionists, with a median wage of $19.00 an hour in that sector (May 2025). Adding a person for the lunch hour or the 7 to 9 a.m. rush is expensive and hard to schedule, and those are the hours that decide whether a new patient books with you or the next practice on the search results.

What an AI receptionist for a medical office should handle

Map your call types before you watch a single demo. The table below is a starting point; your physicians and office manager should adjust it.

Call type	What the AI does	Hand to a person when
Existing patient books, moves or cancels	Verify identity (name plus date of birth), find the visit, offer open slots within your rules, confirm by text	Visit types that need provider approval, or repeated late cancellations
New patient wants an appointment	Collect demographics, reason for visit, referring provider, insurance carrier and member ID; offer new patient slots or create a request	Insurance you do not accept, or a referral-only specialty
Prescription refill	Capture medication, dose, pharmacy and callback number; file a task for clinical staff; give your standard turnaround time	Controlled substances or a caller reporting side effects
Records, forms, referral status	Capture the request and file it in the EHR inbox or ticket queue	Disputes or legal requests
Symptoms or "should I come in?"	Read your fixed safety script and transfer to the nurse line or on-call provider	Always. The AI should not triage or advise
Red flags (chest pain, trouble breathing, stroke signs, suicidal thoughts)	Tell the caller to hang up and call 911 (or 988 for a mental health crisis), as your script states, and alert staff	Always, immediately
Billing and balances	Give payment options and hours, take a message or send a payment link if your system supports it	Disputes, payment plans, collections
Hours, directions, accepted plans	Answer from an approved knowledge base	Anything not in the knowledge base
Pharmacies, labs, other offices	Route to the right extension or take a message	Results or urgent clinical callbacks

The pattern: the AI handles scheduling, intake and messages, and people handle anything clinical, urgent or financial. Our medical practice automation page covers the work around the phone, such as intake forms, eligibility checks and recall.

Packaged product or custom build: which fits your practice

There are three ways to get an AI receptionist into a medical office, and the right one depends mostly on your EHR and how unusual your scheduling rules are.

Question	AI in your phone or patient messaging platform	Standalone healthcare AI receptionist	Custom voice agent
Setup effort	Lowest if you already use the platform	Low to medium; needs an EHR connection to book	Highest; scoped and built for you
Booking into the EHR	Often requests only, unless the platform has a scheduling integration	Varies by vendor and EHR; some book directly, some create tasks	As deep as your EHR's API or partner program allows
Triage and routing rules	Within the vendor's settings	Moderate	Full control by provider, location, visit type, language and on-call rotation
BAA scope	Check whether the AI feature is covered, not only calling	Check the plan tier that includes the BAA	You choose BAA-eligible vendors for voice, telephony and model
Who owns numbers, scripts, data	Platform	Vendor	You, in your own accounts
Best for	Single-location practices already on that platform	Practices that want a packaged product fast	Multi-location groups, mixed EHRs, specialty scheduling

Allo's 2026 comparison of medical AI receptionists makes a useful point about scope: of the five products it reviewed, it found only one whose BAA explicitly covered every feature including the AI, and it could not confirm BAA scope for others. Treat "HIPAA compliant" on a pricing page as a question, not an answer.

Which EHRs an AI receptionist can book into

The voice is the easy part. Whether the AI sees real openings and writes the appointment back is what separates a receptionist from a message taker. How this works differs by system:

• athenahealth. athenaOne exposes a developer API that includes calls to find open appointment slots by department, provider and appointment type, and to book them. That makes direct booking practical for a custom build, subject to athenahealth's API access terms.
• eClinicalWorks, Tebra and NextGen. These support integrations mainly through partner and certification programs rather than open self-serve access, so ask any vendor whether it is an approved integration partner and what it can write, not only read.
• ModMed, AdvancedMD and specialty systems. Integration depth varies widely. Some vendors book directly, some sync on a schedule, and some only create a task for staff.
• No usable API. The fallback is an appointment request queue: the AI collects everything and staff book it each morning. That still recovers after-hours calls, but price it as an answering service.

Questions that expose weak integrations in five minutes:

1. Does it read live availability by provider, location and visit type, or a copy that can double-book?
2. Does it respect visit lengths and rules, such as a 40-minute new patient visit or physicals only on certain days?
3. Can it create a new patient record with insurance fields, or only a note?
4. What happens to calls when the EHR is down?

If you are on Microsoft or Google tools, the request queue can also live in a small internal app; see our AppSheet vs Power Apps comparison.

What an AI receptionist for a medical office costs

Costs come in three parts: a subscription or platform fee, usage (minutes, texts, phone numbers) and the setup work to connect phones and the EHR. Published 2026 prices give a useful range:

• Healthcare-focused products. Allo's 2026 comparison lists Freed's AI receptionist at $149 a month for 250 calls, DeepCura at $129 a month per provider, and Allo's own plan at about $114 a month on annual billing including its BAA fee (Allo is the publisher, so confirm all of these with written quotes).
• Done-for-you services. GetVoIP's 2026 review of HIPAA-compliant AI receptionists lists Smith.ai's done-for-you plans at $500 to $2,000 a month.
• Usage-based platforms for custom builds. Retell AI's pricing page (checked October 2026) lists $0.07 to $0.31 per minute depending on the model and voice, including $0.055 per minute for voice infrastructure and $0.015 per minute for US telephony. Its BAA is only on the Enterprise plan, which a medical build needs.

A worked example. A three-provider family practice sends 600 overflow, lunch and after-hours calls a month to the AI, averaging 3 minutes each: 1,800 minutes.

Setup	Monthly cost	Notes
Custom agent at $0.07 per minute	$126 in usage	Plus a one-time build, numbers and the enterprise BAA tier
Custom agent at $0.31 per minute	$558 in usage	Premium voice and larger model
Per-provider product at $129	$387	Check call caps and what the BAA covers
Done-for-you service	$500 to $2,000	Includes humans on some plans
Staffing 20 more phone hours a week at $19.00	About $1,650 in wages	Before payroll tax, benefits, training and cover for absences

Then run your own number. Export missed and abandoned calls from your phone system, estimate how many were new patients, and multiply by your new patient booking rate and average first-year revenue per new patient. If a handful of recovered new patients a month covers the bill, the case is clear. Our guide to AI automation agency costs explains how one-time build fees are usually priced.

HIPAA, BAAs and calling rules

A medical AI receptionist hears names, dates of birth, symptoms and insurance IDs, all protected health information. This is practical guidance, not legal advice; have your compliance advisor or attorney approve the setup.

• BAA before the first real call. Any vendor that creates, receives, stores or transmits PHI for you is a business associate. Ask whether the BAA covers the AI feature, recordings, transcripts and call summaries, and get the list of subprocessors (telephony, speech-to-text, language model, storage).
• No training on your calls. Get in writing that recordings and transcripts are not used to train shared models.
• Retention and access. Decide how long recordings and transcripts are kept, who can listen, and whether summaries go into the chart. Less stored is less risk.
• Identity checks. Require at least name and date of birth before discussing any appointment, and never read back results or diagnoses.
• Recording consent and AI disclosure. Some states, including California, require all parties to consent to recording. A greeting that says the caller is speaking with an automated assistant and the call may be recorded covers both.
• Outbound calls. On 8 February 2024 the FCC ruled that AI-generated voices count as "artificial" under the TCPA, so outbound AI reminder or recall calls need the same consent as prerecorded calls. Keep appointment reminders separate from marketing and honor opt-outs.

How to set up an AI receptionist in a medical office, step by step

1. Pull 60 to 90 days of call data. Calls by hour and day, missed and abandoned calls, voicemails and the reasons staff log. Mondays, opening hour and lunch are the usual gaps.
2. Map call types and rules. Use the table above. Add visit types and lengths, which providers see which visit types, accepted insurance plans, locations and languages.
3. Write the red-flag list and safety script with your physicians. The clinicians approve the exact wording, including when to say 911 or 988 and how to reach the on-call provider.
4. Sign BAAs and connect the EHR. Use a dedicated service account with the least access needed. Confirm whether booking is live or request-based.
5. Choose the routing. Start narrow: after hours only, or overflow after a set number of rings. Widen once results are proven.
6. Test with real scenarios. Replay 30 to 50 past calls, including elderly callers, poor lines, Spanish speakers if you serve them, angry callers and every red-flag phrase. Fix each wrong outcome.
7. Run in shadow mode. For the first two weeks, staff review every AI booking, request and escalation each morning.
8. Measure and tune. Track answered, booked, requested, escalated and abandoned calls by hour and location, and review red-flag transfers weekly.

A platform add-on can be switched on in days, though testing still applies. A custom agent with EHR booking, escalation and reporting typically takes 4 to 8 weeks in stages.

Common mistakes, and when not to buy

• Trialing on real patients without a BAA. A free trial on a non-BAA plan is a compliance problem from the first call.
• Letting the AI answer clinical questions. Even "is this normal after my procedure?" goes to a nurse.
• Expecting it to fix eligibility and prior auth. Those were the biggest phone burden in the MGMA poll, but they are back-office work. Eligibility checks and document processing automation for referrals and auth paperwork address them better than a voice agent.
• No one reads the request queue. Every unbooked call needs a named owner and a morning callback.
• Measuring calls answered instead of visits kept. Track booked and completed visits from AI-handled calls.

Skip it or wait if your team already answers nearly every call within a few rings, your call volume is low, your EHR schedule templates are messy, or most of your calls are clinical (for example, a high-acuity specialty where callers mainly describe symptoms). Fix the templates first, or start with reminders and online booking.

Getting an AI receptionist built for your medical office

If your phone or patient messaging platform's AI books into your EHR, covers the AI under its BAA and follows your red-flag rules, use it. Bring in an agency when you run several locations or EHRs, need custom routing by provider and visit type, want refill and records requests filed as EHR tasks, or want call outcomes in one dashboard.

We build AI receptionists and AI voice agents for medical practices, connected to the EHR and phone system you already run. Every project starts with a free review of your call data, gets a fixed price, ships in stages and runs on accounts you own, with BAAs in place before go-live. Call outcomes can feed a Looker Studio dashboard. For other practice types, see our AI receptionist for dental office guide, and for picking any partner, how to choose an AI automation agency.

To get started, book a free review and tell us which EHR and phone system you use and roughly how many calls go unanswered each week.

FAQ

Frequently asked questions