Quick answer
Before you buy an AI receptionist, ask how it hands a caller to a person, what it does after hours, which calendar and records it writes to, how it's told your prices and policies, who can hear the recordings, and how it tells callers it's an automated assistant. Then run it on a test line with your own call scripts before you sign.
Key takeaways
- Ask for a live test call with your own scripts, not only a recorded demo.
- The handoff to a person matters more than the voice.
- Ask who can hear recordings, how long they're kept and whether they train the vendor's models.
- Have it say at the start of every call that it's an automated assistant, so callers know what to expect.
- The 30-minute assessment reviews your call volume and missed calls before any purchase.
Start your 30-minute AI assessment.
What should an AI receptionist do on a typical call?
It should answer within a few rings, find out what the caller needs, answer routine questions from your written policies, collect the caller's name, number and reason, book or request an appointment, and send a summary to your team. When the caller is upset, confused or asks for a person, it should transfer or promise a callback.
Write down your five most common call types before you talk to any vendor. For each one, note what a good receptionist says and what has to be captured. A plumber's list might be: emergency leak, estimate request, existing job question, invoice question, and sales calls. A dental office's list looks different.
Then ask each vendor to handle those five calls, in your words, on a live test line. A polished demo of someone else's business tells you little.
Worked example (hypothetical, not a client case). An Anaheim garage door company tests two AI phone services with the same five scripts. One books the estimate correctly but can't transfer a caller who asks for the owner. The other transfers cleanly but takes three tries to capture the address. The owner picks the second and adds an address check to its script. No results are claimed.
Which questions should you ask an AI receptionist vendor?
Ask how it hands off to a person, what happens after hours, which calendar and records it writes to, how you update prices and policies, how it handles a caller it cannot understand, who can hear recordings and transcripts, how long they are kept, whether your data trains their models, what it costs per minute, and how you cancel.
Put the questions in writing and keep the answers:
- Handoff. How does a caller reach a person, and what happens if nobody picks up?
- After hours. Does it book, take a message or page someone on call?
- Systems. Which calendar, CRM or job system does it write to, and what happens when that system is down?
- Knowledge. How do you change a price, a service area or a holiday schedule, and how fast does the change take effect?
- Hard calls. What does it do with a strong accent, a bad connection or a caller who won't answer questions?
- Recordings. Who can listen, where are they stored and how long are they kept?
- Training. Is your call data used to train the vendor's models?
- Disclosure. What does it say at the start of the call?
- Pricing. Per minute, per call or flat, and what counts as a call?
- Exit. How do you cancel and get your recordings and contacts back?
For a trades firm, the home services post describes what booking and dispatch should look like once the system is live.
What should an AI receptionist say at the start of a call?
It should greet the caller with your business name, say it's the automated assistant, and tell callers if the call is recorded or transcribed. Then it should say what it can help with and how to reach a person. A clear opening sets expectations, cuts frustration and makes the handoff feel normal instead of like a failure.
A simple opening works: "Thanks for calling [company]. You're speaking with our automated assistant, and this call may be recorded or transcribed. I can book a visit, answer common questions or get you to a team member. How can I help?"
Ask each vendor to show you its default greeting, and change it to your own words before launch. Then check three things on a test call:
- Clarity. Would a first-time caller understand in one listen that they're talking to an automated system?
- Length. The opening should take under ten seconds. Long greetings make people hang up.
- The way out. Saying "person" or "representative" at any point should reach a human or a callback promise.
Outbound calling is a different project. If a vendor offers AI callbacks or outbound campaigns, treat them as a separate decision with its own testing, and start with inbound calls only.
How do you test an AI receptionist before going live?
Run it on a test number for one to two weeks with staff playing callers. Use your five common call types, plus hard cases such as an angry caller, a wrong number and a caller who asks for a person. Read every summary. Go live on after-hours calls first, then add daytime overflow once the summaries are reliable.
Score each test call on four points: did it understand the request, did it capture the right details, did it book or route correctly, and did the summary match the call. Keep the scores in a shared sheet so you can compare vendors or settings.
After hours is the safest place to start, because the alternative is usually voicemail. Daytime overflow, meaning calls your front desk can't reach, is the next step. Answering every call comes last, if at all.
For a practice that answers patient calls, the dental and med spa post covers scheduling and the questions an AI should never answer.
How do you tell whether an AI receptionist is worth the cost?
Count missed calls and voicemails for two weeks before you start, then compare after 60 days. Also track booked appointments from calls, the share of calls handed to a person, and complaints about the phone experience. If booked work rises and complaints stay flat, the system is probably paying for itself.
Your phone provider can usually export call logs with answered, missed and voicemail counts. Pull two normal weeks before the AI starts.
Then compare. Missed calls should fall. Booked appointments from phone calls should hold or rise. A high handoff rate in the first month is normal while the scripts are tuned. A rising complaint count is the warning sign. See How do you measure AI ROI in a small business? for a simple way to put a dollar figure on the change.
Who helps Orange County businesses choose an AI receptionist?
Preferred AI Partners is a fractional AI partner in Irvine that works with businesses of 10 to 200 people. It reviews your call logs, writes the call scripts with you, tests the options against them and sets up the handoff to your team. The 30-minute assessment starts with your missed calls.
For how AI agents fit together across a business, read What can AI agents do for a small business? The Orange County page and the services page describe the work. Ask any vendor for a live test line before you sign.
