AI Receptionist for Small Business: What It Does, What It Costs, What the Law Says

We make Extra, an AI agent platform, so this article is written by an interested party. Facts about other tools and about the law come from the sources listed at the bottom, as of the date shown.

What is an AI receptionist?

An AI receptionist (also sold as an AI answering service or virtual receptionist) is a voice agent connected to your business phone line. It picks up, understands what the caller says, answers out loud and takes action in your tools. Unlike voicemail, it holds a real conversation, based on the instructions and answers you give it.

A good AI receptionist does four things:

• Answers common questions: hours, service area, what a standard job costs, what to bring to an appointment.
• Books appointments in your calendar, in the time slots you have opened, without double-booking.
• Takes messages and sends you a short summary of every call, with the next step.
• Puts urgent calls through to you right away.

That is what Hugo, the phone agent from Extra (hello-extra.com), does: he answers your calls, books appointments in your Google Calendar, sends you a summary after each call and passes urgent calls to you. You write his greeting and his answers with him. Details are on Hugo’s page.

What does an AI receptionist handle well, and what does it handle badly?

An AI receptionist is strong on repetitive, predictable calls and weak on emotional or unscripted ones. Before you set one up, look at what kind of calls fill your week.

What it handles well

• Questions that keep coming back. A landscaping company gets the same calls every spring: “Do you cover my town?”, “When can you start weekly mowing?” The agent answers the fiftieth call as patiently as the first.
• Routine scheduling. A dental office that opens cleaning and checkup slots can let the agent offer and book them.
• Calls nobody can take. The HVAC tech in an attic, the front desk on a busy morning: the call is answered and summarized, even evenings and weekends if you choose.

What it handles badly

• Emergencies described vaguely. A homeowner who says “the furnace is acting weird” may have no heat with a newborn at home, or smell gas. The agent only escalates what you taught it to recognize as urgent. Vague rules, wrong calls.
• Angry callers. An upset customer wants a human first. The agent can log the complaint, but it won’t calm things down the way you would. A sound rule: every complaint gets transferred, or a same-day callback from you.
• Strong accents, noise, bad lines. Speech recognition is not equally accurate for everyone. A Stanford-led study published in PNAS in 2020 tested five commercial systems (Amazon, Apple, Google, IBM, Microsoft) and found an average word error rate of 0.35 for Black speakers versus 0.19 for white speakers. The systems have improved since; still, test with voices that sound like your customers.
• Transfers when you can’t pick up. Transferring an emergency only works if someone answers. Set a backup number, or a clear promise: “I’m alerting the owner, he’ll call you back within the hour.”
• Complex requests. A system replacement quote, a billing dispute, a clinical question: the agent should take details and hand off, not improvise.

When not to use one: if most of your calls are true emergencies or negotiations, keep a human on the first ring and use AI for overflow and after hours.

How much does a virtual receptionist cost compared with an answering service or a part-time hire?

For a small business taking around 100 calls a month, an AI receptionist usually costs less than a live answering service and far less than a part-time employee. Live receptionists still handle delicate calls better. Here are the numbers, as of October 2026.

A few things to keep in mind:

• Minutes and calls are not the same unit. At two to three minutes a call, Ruby’s 200 minutes cover roughly 70 to 100 calls.
• A part-time hire only covers their own shift, but they also greet walk-ins, chase paperwork and handle billing.
• At Extra, the $99 plan also includes the seven other agents (marketing, website, prospecting, finance and more). Unused credits stay valid for 60 days. See the pricing page for details.

What are the rules for an AI receptionist in the US?

Three sets of rules matter: state call-recording laws, the TCPA for any outbound calls, and HIPAA if you handle patient information. Telling callers they’re talking to AI is, in any case, good practice.

Call recording: one-party and all-party consent states

Federal law lets a call be recorded with the consent of one party, but about a dozen states require every party to agree, at least in some situations. They include California, Florida, Illinois, Maryland, Massachusetts, Montana, Nevada, New Hampshire, Pennsylvania and Washington, with nuances in Connecticut, Delaware and Michigan.

California is the strictest well-known example. Penal Code section 632 bars recording a confidential conversation without the consent of all parties, with a fine of up to $2,500 per violation for a first offense. According to the Reporters Committee for Freedom of the Press, calls on cell or cordless phones need everyone’s consent whether or not they are confidential. Your callers may phone in from any state, so the simple habit is to say at the start of the call that it may be recorded or transcribed.

The TCPA and AI voices on outbound calls

On February 8, 2024, the FCC announced a unanimous declaratory ruling: calls made with AI-generated voices count as “artificial” under the Telephone Consumer Protection Act. That means the TCPA’s consent rules for artificial or prerecorded voice calls apply, including prior express written consent for telemarketing robocalls. The ruling targets calls you place. An AI receptionist that answers calls customers make to you is a different situation, but the moment you use an AI voice to call leads or send reminders, the TCPA applies. If outbound is on your list, read our article on AI prospecting first.

HIPAA for dental and medical offices

A dental office whose AI receptionist hears patient names, symptoms and appointment details is sharing protected health information with a vendor. Under HIPAA, a vendor that handles that information for you is a business associate, and you need a business associate agreement (BAA) with it before sharing. Ask any AI receptionist vendor about this before you connect it.

Data: know where it goes

Ask every vendor where call data is stored, who processes it and how long transcripts are kept. Keep the summaries you need and delete the rest. At Extra, data is stored with Cloudflare and Supabase, mostly in Western Europe, and request content is processed by AI models in the United States; the security page explains how it is protected.

Disclosure: say it’s AI

Even where no rule forces it, tell callers up front that they’re speaking with an AI. Someone who finds out three minutes in feels tricked, and that feeling ends up in a review. Extra’s rule is simple: Hugo says he’s an AI at the start of every call.

How do you set up an AI receptionist? A checklist

Plan on half a day of setup. Most failures come from rushed instructions, not from the technology.

1. List your 10 to 15 most frequent questions (track them for a week), with the exact answer you want callers to hear.
2. Write the greeting: business name, the fact that it’s an AI, what it can do. Two sentences.
3. Define emergencies with concrete examples (“no heat and it’s below freezing,” “gas smell,” “broken tooth with swelling”) and who gets the transfer.
4. Plan the fallback when nobody picks up a transfer: a second number, or a callback time you actually keep.
5. Open your calendar slots: appointment length, drive time, closed days, maximum bookings a day.
6. Set what the agent never does: no quotes on custom jobs, no medical or legal advice, no promised dates without you.
7. Choose coverage: every call, or only overflow, lunch, evenings and weekends.
8. Place at least ten test calls: a rushed customer, an angry one, a real emergency, an off-topic question, an accented voice, a call from a noisy street. Fix the answers after each one.
9. Read every summary in the first week: that’s where the missing questions show up.

How do you tell callers an AI answers the phone?

Say it in the first sentence and always offer a way to reach a person. Here is a sample greeting for a dental office:

“Thanks for calling Bright Smile Dental. I’m the office’s AI assistant: I can answer questions, book a visit or take a message. If you’re in pain or it’s urgent, tell me and I’ll get Dr. Patel right away.”

• Tell regular customers by email or on your Google Business Profile.
• Don’t pass the agent off as an employee. A first name is fine; a made-up human identity is not.
• Call back fast whenever the agent promised a callback. Trust is won or lost there.
• Watch your reviews: an annoyed caller will say so on Google. Our guide on replying to Google reviews can help.

Is an AI receptionist right for your small business?

An AI receptionist pays off if you miss calls because your hands are busy and most of those calls are questions or bookings: a dental office, a landscaping company, a salon, a solo contractor on a job site. For emergency HVAC or plumbing service, keep a human on call behind it. If every call is a negotiation or a crisis, skip it.

To see where an AI receptionist fits among other tools, read what an AI agent is for a small business and AI agent vs chatbot vs automation.

Also read: 12 business phone greeting scripts.

Also read: AI appointment booking and AI receptionists for dental offices.

Sources

1. Ruby, “Pricing,” accessed 10/25/2026
2. Smith.ai, “Pricing,” accessed 10/25/2026
3. Smith.ai, “AI Receptionist Pricing,” accessed 10/25/2026
4. U.S. Bureau of Labor Statistics, “Receptionists,” Occupational Outlook Handbook, accessed 10/25/2026
5. FCC, “FCC Makes AI-Generated Voices in Robocalls Illegal,” 02/08/2024
6. Justia, “Recording Phone Calls and Conversations: 50-State Survey,” accessed 10/25/2026
7. Justia, “California Penal Code Section 632,” accessed 10/25/2026
8. Reporters Committee for Freedom of the Press, “Reporters Recording Guide: California,” 01/08/2026
9. U.S. Department of Health and Human Services, “Business Associates,” 10/21/2026
10. PNAS (Koenecke and coauthors), “Racial disparities in automated speech recognition,” 04/07/2020