A bilingual AI receptionist for a medical practice answers patient calls in more than one language, usually English and Spanish in the US, and completes routine requests such as booking and refill intake in the caller's language. It is only worth having if the caller can finish the whole call in that language and reach a person when the call goes beyond routine. Plenty of products can say hola. Far fewer can get a date of birth, a medication name and a pharmacy right in Spanish and write them correctly to the chart.
A bilingual AI receptionist is an AI phone agent that holds patient calls in more than one language and completes the same administrative tasks in each: patient identification, scheduling, refill request intake and routing to staff. It is an administrative tool. It is not an interpreter, and it should not be used where a qualified interpreter is required.
You will not find a list of the languages MedPhone handles in this post. Language capability in this category changes faster than a blog post does, and the honest way to answer the question is on a demo, with your own test calls. Ask us what is live, then make us prove it.
Why a bilingual AI receptionist matters to a medical practice
Roughly one in five people in the US speak a language other than English at home, according to the Census Bureau, and at some practices that describes most of the panel. A caller with limited English proficiency who cannot make themselves understood on the first try has little reason to try again tomorrow.
The usual answer is a Spanish speaking receptionist, if you can hire one, and a telephone interpreter service for everyone else. Both work. The first does not stretch past one person and one shift. The second adds a wait and a three way call to what should be a two minute reschedule. A phone agent that handles routine calls in the caller's language can take some of that load, provided it actually works for your multilingual patients.
Language access rules: Section 1557 and your phone line
I am not a lawyer and this is not legal advice, but every practice manager should know the outline.
Section 1557 of the Affordable Care Act prohibits discrimination on the basis of national origin, among other grounds, in health programs that receive federal financial assistance. The HHS rule that implements it, 45 CFR Part 92, requires covered entities to take reasonable steps to provide meaningful access to people with limited English proficiency. In practice that means free language assistance, qualified interpreters when interpreting is needed, not relying on patients to bring their own interpreter, and not using a child to interpret except in an emergency.
Two points matter for anyone putting AI on the phone line.
- An AI agent is not a qualified interpreter. The rule was written with people in mind, and how it applies to software holding a conversation directly in a patient's language is not settled. Treat the agent as an administrative channel, not as your language access plan
- The 2024 rule says machine translation of critical content needs review by a qualified human translator. That provision is about written translation, but it is a fair signal of how regulators view unsupervised machine language in health care
The safe line is simple. Administrative calls in the caller's language are a convenience. Anything clinical, anything about consent, and anything where a misunderstanding could harm the patient goes to a person, with a qualified interpreter where one is needed.
Spanish speaking receptionist, interpreter line or AI?
| Option | Covers | Weak spot | Fit for clinical calls |
|---|---|---|---|
| Bilingual front desk staff | One language, opening hours | One person, one shift, hard to hire | Only if trained and qualified |
| Telephone interpreter service | Many languages, any hour | Wait to connect, three way call | Yes, with qualified interpreters |
| AI receptionist in the caller's language | Routine administrative calls | Quality varies by language and accent | No, should transfer |
| English only line plus voicemail | Nothing, for these callers | Callers hang up and do not return | No |
Practices with a large Spanish speaking panel usually end up with a mix: bilingual staff or an AI agent for scheduling and refill intake, and an interpreter service for everything clinical.
How do you test an AI receptionist in Spanish?
With live calls you place yourself, using native speakers from your own staff or patient advisory group, ideally people who speak the regional variety your patients speak. The same eight tests work for any language.
- Speak only Spanish from the first word. Does it switch, or does it make the caller press a button or ask for Spanish in English first?
- Switch languages halfway through, the way bilingual callers really do. Does it follow, or lose the thread and start over?
- Give a date of birth and a phone number in Spanish, fast, the way patients do. Then check what it wrote to the chart
- Spell a surname common in your panel, and ask for a provider whose name is not Spanish. Both should come out right
- Request a refill for a medication that has both a brand and a generic name, and name the pharmacy. Check the task in the provider inbox
- Describe a symptom partway through. It should stop and route the call, in the same language, to the right place
- Ask for a person. It should transfer, and you should find out what language the person who picks up speaks
- Listen to the confirmation. A booking read back in English to a Spanish speaking caller is a booking the caller may not have understood
Then ask the vendor three questions in writing. Which languages are live in production today, on your EHR, rather than on the roadmap? What is containment in that language on live deployments, compared with English? And in which language is the call summary written, and can your staff read it?
Where the chart and the staff fit in
A call in a second language still creates a record your English speaking staff have to act on. The call summary in the chart, the refill task in the provider inbox and the reason for a cancellation all need to be readable by whoever picks them up. Ask to see a real summary from a second language call before you sign.
The transfer path matters as much. If the agent hands a Spanish speaking caller to a front desk that only speaks English, the problem has moved, not gone away. Decide in advance where those transfers go: a bilingual staff member, a direct line to your interpreter service, or a callback from someone who speaks the language, within a stated time.
What MedPhone does on a call, and what to ask about language
The parts of MedPhone that do not depend on language I can state plainly. It identifies the patient against the EHR, books, reschedules and cancels against real availability, takes refill requests into the provider inbox for a clinician to approve, writes a call summary to the chart, and transfers clinical, urgent or unusual calls to staff under rules the practice sets. It gives no clinical advice, and it is not an interpreter.
For language, ask us on a demo what is live in production on your EHR, and run the eight tests above on us exactly as you would on anyone else. If it does not pass them for your patients, it is the wrong tool for that part of your phone line, and you should know that before you buy.
See how a call works end to end
Sources
Questions people ask
What is a bilingual AI receptionist for a medical practice?
It is an AI phone agent that holds patient calls in more than one language and completes routine administrative tasks in each, such as scheduling and refill request intake. It is not an interpreter and should transfer clinical calls to a person.
Does an AI receptionist meet Section 1557 language access rules?
Not on its own, and you should not rely on it to. Section 1557 requires meaningful access for people with limited English proficiency, including qualified interpreters where interpreting is needed, and how the rule applies to AI agents is not settled. Treat the agent as a convenience for administrative calls and keep a qualified interpreter path for everything else.
Does MedPhone speak Spanish?
Ask us on a demo what is live in production on your EHR, then test it with native speakers using the checklist in this post. Language capability changes faster than a blog post, so judge it on your own calls rather than on a sentence here.
How do I test an AI receptionist in Spanish?
Use native speakers, speak only Spanish from the first word, switch languages mid call, and give a date of birth and medication name quickly. Then check what was written to the chart and the provider inbox, and test what happens when the caller asks for a person or mentions a symptom.
Can a bilingual AI receptionist answer clinical questions?
No. No AI receptionist should give clinical advice in any language. Clinical calls should transfer to clinical staff, with a qualified interpreter where the patient needs one.
Should a clinic hire a Spanish speaking receptionist instead?
If you can hire and keep one, a bilingual person handles difficult and clinical conversations better than software. The limit is coverage, since one person covers one shift and one language. Many practices use bilingual staff for complex calls and an AI agent for routine calls at the 8am rush and after hours.
Written by
Agni Patel is the founder and CEO of MedPhone. This post is about how to evaluate any vendor, his included, and it deliberately does not list what languages MedPhone supports.