AI Receptionists and Voice Agents: What They Can and Can't Do

A missed call can mean a missed job. That simple fact is why so many small businesses are looking at AI receptionists, also called voice agents: software that answers the phone, talks with callers in everyday language, and takes some action on the business's behalf.

For routine calls, these tools can do real work. They also have real limits, and a few legal lines matter more than most sales pitches let on. This guide covers what a voice agent can reasonably handle, where it struggles, how to design the handoff to a person, and what federal regulators have said about AI-generated voices on the phone.

This is general information, not legal advice. Rules on calling, recording and disclosure depend on the situation and the state, so talk with a qualified attorney before you launch.

What a voice agent can do well

Voice agents do best with calls that are frequent, predictable and well documented. Typical jobs include:

• Answering every call, including after hours, during lunch, and when your team is already on the line.
• Answering common questions from information you have approved: hours, service area, published prices, directions, what to bring to an appointment.
• Booking, rescheduling and canceling appointments when connected to your calendar with clear rules about what can be booked and when.
• Taking intake details such as name, callback number, address and the reason for the call, then writing a summary into your customer records.
• Routing: sorting sales, billing, support and urgent calls, then transferring the caller or taking a detailed message.
• Sending a confirmation text afterward, where the caller has agreed to receive one.

For example, imagine a three-person HVAC company. At 9 p.m. a caller says the heat is out. The agent confirms the address, offers the next open slot, and, because the owner wrote a rule that "no heat with an infant or elderly person at home" is urgent, pages the on-call technician instead of waiting for morning. The agent did not decide what counts as urgent; the owner did.

Where voice agents fall short

The limits matter as much as the features.

• Confident mistakes. Generative AI can state wrong information fluently. NIST's Generative AI Profile calls this confabulation, "the production of confidently stated but erroneous or false content." On the phone, that might be a price you do not offer or a same-day visit you cannot make.
• Complex or emotional calls. In its June 2023 report on chatbots in consumer finance, the Consumer Financial Protection Bureau (CFPB) concluded that chatbots "may be useful for resolving basic inquiries, but their effectiveness wanes as problems become more complex." It also described "doom loops," where customers get stuck in repetitive responses "without an offramp to a human customer service representative."
• Judgment calls. Refunds, exceptions, billing disputes, complaints, and anything involving health, safety or legal questions belong with a person.
• What it can see. An agent can only book what your calendar shows and quote what you gave it. Out-of-date information produces out-of-date answers.
• Real-world audio. Test with speakerphones, noisy job sites, weak cell connections and callers who interrupt, not just a quiet office.

Design the handoff before you go live

The single most important design decision is when the agent stops and a person takes over. Write the rules down before launch:

• The caller asks for a person. Honor it promptly.
• The agent misunderstands the caller twice in a row.
• The call involves an emergency, a complaint, a cancellation of a large job, or a topic on your "humans only" list.
• The agent cannot find the answer in its approved information.

Then decide how the handoff happens. A warm transfer, where the agent passes a short summary along with the call, spares the caller from repeating everything. When nobody is available, the agent should take a complete message and give a realistic callback time, and someone should own that callback queue. The guide "Human in the Loop: Keeping People in Charge of AI Agents" covers escalation and takeover in more depth.

Disclosure: telling callers they are talking to AI

Federal rules treat outbound and inbound calls differently.

Outbound calls. In February 2024 the Federal Communications Commission (FCC) issued a declaratory ruling confirming that the Telephone Consumer Protection Act's restrictions on "artificial or prerecorded voice" encompass current AI technologies that generate human voices. As a result, those calls require the prior express consent of the called party, absent an emergency purpose or exemption. The FCC's announcement notes that its rules require telemarketers to obtain prior express written consent before robocalling consumers, and the ruling points out that artificial or prerecorded voice messages must identify the entity responsible for the call. If you want a voice agent to place reminder, follow-up or sales calls, get the consent question settled first.

Inbound calls. In an August 2024 proposal, the FCC wrote that "the TCPA's requirements do not extend to technologies used to answer inbound calls." That same proposal would require callers using AI-generated voice to disclose, at the beginning of each call, that the call uses AI-generated technology. It was a proposal, not a final rule, so check its current status.

Even where no rule requires it, a short, plain disclosure is good practice: "Thanks for calling [business name]. I'm the office's automated assistant. I can book or change appointments, or connect you with the front desk." It sets expectations and gives the caller an easy way out.

Do not let the agent claim to be human. The Federal Trade Commission (FTC) describes an ad as deceptive when it contains "a statement - or omits information - that is likely to mislead consumers acting reasonably under the circumstances" and that information is material. A business that tells customers they are talking to a person when they are not is inviting exactly that kind of problem, and it erodes trust when callers figure it out.

If the agent records or transcribes calls, say so at the start. Consent rules for recording are set largely at the state level, so confirm what applies to you and your callers.

Buying carefully: test the promises

Be wary of sweeping claims. In August 2025 the FTC sued a company that marketed "conversational AI" it said could replace human customer service representatives, alleging deceptive earnings promises and refund guarantees. That case was about business-opportunity claims, but the lesson carries over: ask for evidence, not adjectives.

Before you sign, ask any provider:

1. Can I hear real, unedited call recordings, with permission, from businesses like mine?
2. What happens when the agent does not know an answer?
3. How do callers reach a person, and how fast?
4. Can I see and correct what the agent says, and review transcripts?
5. Where are recordings and transcripts stored, and for how long?
6. Is there a trial period, and how do I turn it off quickly?

Key takeaways

• Voice agents handle routine, well-documented calls best: answering, FAQs, booking, intake and routing.
• They can state wrong information confidently, so limit them to approved information and test with real calls.
• Write escalation rules before launch and make it easy for callers to reach a person.
• AI-generated voices on outbound calls fall under the TCPA's artificial-voice rules and need prior express consent, absent an exemption.
• Disclose that callers are talking to an automated assistant, and never let it claim to be human.
• This is general information, not legal advice; confirm the rules that apply to your calls.

Sources

• Federal Communications Commission, FCC Makes AI-Generated Voices in Robocalls Illegal (February 8, 2024)
• Federal Communications Commission, Declaratory Ruling, FCC 24-17, CG Docket No. 23-362 (February 8, 2024)
• Federal Communications Commission, Notice of Proposed Rulemaking and Notice of Inquiry, FCC 24-84 (August 8, 2024)
• Consumer Financial Protection Bureau, Chatbots in consumer finance (June 2023)
• National Institute of Standards and Technology, NIST AI 600-1, Generative Artificial Intelligence Profile (July 2024)
• Federal Trade Commission, Advertising FAQs: A Guide for Small Business
• Federal Trade Commission, FTC Sues to Stop Air AI from Using Deceptive Claims about Business Growth, Earnings Potential, and Refund Guarantees to Bilk Millions from Small Businesses (August 25, 2025)