What does an AI receptionist or website chat assistant actually do?

The four jobs

1. Answer. It responds to the questions you hear over and over: hours, location, services, how booking works, what to bring. It should answer from a set of approved information you provide, such as your hours, service list and policies, not from whatever it guesses. Lesson 3 explained why: AI can produce confident answers that are wrong when it has to fill a gap [1].

2. Capture. It collects the details a person on your staff would ask for: name, best phone number or email, what they need and when. Those details go somewhere a person will follow up, such as your email or scheduling system.

3. Book. Where it is connected to your calendar or booking system, it can offer open times and set an appointment. Where it is not, it can collect preferred times and pass them along.

4. Hand off. It knows when to step aside: when someone asks for a person, when the question is outside its approved information, when the caller is upset, or when the topic involves money, a complaint or anything sensitive. A good hand-off passes along what was already said.

What it should not do

Lesson 5 covered keeping decisions with people. For an assistant that talks to customers, that means:

• No promises about prices, discounts, refunds or exceptions beyond your approved information.
• No legal, tax or financial advice.
• No pretending to be a person.
• No collecting sensitive data it does not need, such as full card numbers or ID numbers. (Lesson 4 covers why.)

Say that it is AI

People should know when they are talking with an AI assistant. A simple greeting does it: "Hi, I'm the virtual assistant for Lakeside Landscaping. I can answer questions and help you book. If you'd like a person, just say so."

This is about trust. NIST, the U.S. standards agency, notes that a language model can falsely claim to be human, which can lead people to believe they are speaking with another person [1]. Being upfront avoids that. Honesty rules still apply, too: the Federal Trade Commission's advertising guidance says advertising must be truthful and non-deceptive [2]. A sensible rule of thumb is to hold anything your assistant tells customers to the same honesty standard as anything else your business says. Disclosure rules can also vary by state, so check with an attorney about the rules where you operate.

A note on outbound calls

Answering calls that come to you is different from placing calls to people. In a declaratory ruling released February 8, 2024, the Federal Communications Commission confirmed that AI-generated voices count as "artificial" voices under the Telephone Consumer Protection Act, the main federal law limiting robocalls [3][4]. In practice, that means calls using an AI-generated voice generally require the called person's prior express consent, absent an emergency or an exemption, and calls that contain advertising or telemarketing require prior express written consent [4]. The FCC's rules also require such calls to identify the business responsible [4].

If you ever want an AI voice to make outgoing calls, such as reminders or follow-ups, get legal advice first and make sure your consent records are in order. This lesson is general education, not legal advice.

How to set one up well

• Write the approved answers first. List your top twenty questions and the exact answers. This becomes the assistant's knowledge.
• Set the limits in writing. What it must never say or promise, and when it must hand off.
• Decide where details go. Every captured lead should land somewhere a person checks daily.
• Test it yourself. Call or chat as a customer would. Ask easy questions, odd questions and questions it should refuse.
• Read the transcripts. For the first few weeks, review conversations regularly and fix wrong or awkward answers.

Common mistakes

No hand-off path. If the assistant cannot reach a person, frustrated customers have nowhere to go.

Letting it improvise. An assistant without approved information will fill gaps with guesses.

Hiding that it is AI. It puts trust at risk the moment a customer realizes.

Set and forget. Hours, services and prices change. If the assistant's information does not, it gives outdated answers.

Worked example

A car dealership sets up a website chat assistant for after-hours visitors.

Its approved information covers sales and service hours, the address, how to book a service appointment, and which documents to bring for a test drive. Its limits: no pricing, no financing terms, no trade-in values, and a hand-off for any of those.

At 9 PM a visitor asks whether a particular SUV is still available and what the monthly payment would be. The assistant replies: "I'm the dealership's virtual assistant. I can't quote prices or payments, but I can have a sales consultant reach out first thing tomorrow. What's the best way to contact you, and is there a time that works?" It captures the name, phone number and vehicle of interest, and the lead lands in the sales manager's morning list with the full chat attached.

The visitor got a fast, honest answer, and pricing stayed with a person.

Checklist

• The assistant introduces itself as an AI assistant.
• It answers only from approved information I wrote.
• It has written limits on prices, promises and advice.
• It hands off to a person on request and for sensitive topics.
• Captured details go somewhere a person checks every day.
• I review transcripts and update its information regularly.
• I get legal advice before using an AI voice for outbound calls.

Try it

Write down the ten questions your business answers most often by phone or message. Write the exact answer you would want given for each. Then list five topics an assistant should never answer on its own. Those two lists are the foundation of any AI receptionist or chat assistant.

Tips

• Start the conversation by saying it is an AI assistant and how to reach a person.
• Keep its answers short and tied to your approved information.
• Pass the full conversation along at hand-off so customers do not repeat themselves.
• Update the assistant whenever your hours, services or policies change.

Sources

1. National Institute of Standards and Technology, "Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile," NIST AI 600-1, July 2024, https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf
2. Federal Trade Commission, "Advertising FAQ's," FTC business guidance, https://www.ftc.gov/business-guidance/resources/advertising-faqs-guide-small-business
3. Federal Communications Commission, "FCC Makes AI-Generated Voices in Robocalls Illegal," news release, February 8, 2024, https://www.fcc.gov/document/fcc-makes-ai-generated-voices-robocalls-illegal
4. Federal Communications Commission, Declaratory Ruling, FCC 24-17, CG Docket No. 23-362, adopted February 2, 2024, released February 8, 2024, https://docs.fcc.gov/public/attachments/FCC-24-17A1.txt