What Information Does an AI Receptionist Need Before It Goes Live?

A junk removal owner in Spring finishes a job at 4 pm, sits in the truck, and opens the email from his new AI receptionist provider. "Please send us your information." He pastes his website link and hits reply. Two weeks later, a caller asks whether he hauls hot tubs, and the AI doesn't know. He does. He just never wrote it down.

That gap is what onboarding is for.

What Does AI Receptionist Setup Actually Involve?

Setting up an AI receptionist means documenting how your business answers the phone. You provide your hours, service area, services, common questions and answers, calendar and booking rules, who gets transferred calls, and what the AI shouldn't discuss. The AI uses that knowledge base to answer, book, text, and transfer.

The voice and greeting take minutes. The knowledge takes a few focused hours.

Why Is the Real Work in the Details?

The first AI receptionist prospect we're working with is a pest control company. Their first two requirements had nothing to do with the voice or the script. They needed the AI to work with GorillaDesk, the software they schedule in, and they needed the right calls forwarded to the owner.

That's typical. When I set up FoundOnline's own business line, the hard questions were the same: where does a booking go, and who gets the transfer?

What surprised me is how much a business knows that isn't written anywhere. Owners answer the same questions a dozen times a week and have never typed out the answers. The AI can only say what you've told it.

What Does the Research Say About Getting This Right?

Setup quality matters because first impressions on the phone carry a lot of weight.

• Small-business call studies put unanswered inbound calls at roughly 40% to over 60%. An AI receptionist fixes that, but only if the calls it answers go well.
• A commonly cited industry figure says about 85% of callers who don't get through won't call back. A caller who gets a confused answer is in nearly the same spot as one who got voicemail.

A couple of compliance items belong on the setup list too:

• Call recording: Texas is a one-party consent state, while California and some other states require all-party consent. If you record calls, add a short recording disclosure no matter where you are.
• Healthcare: HIPAA applies to patient call data. Use only a vendor that will sign a Business Associate Agreement.

What Does a Thin Knowledge Base Cost?

Here's a worked example. Use your own numbers.

Say an AI receptionist answers 150 calls a month. If 10% of those calls ask something the knowledge base doesn't cover, that's 15 callers who hear "I'll have someone get back to you" instead of an answer or a booking.

If half of them move on to a competitor while they wait, that's 7 or 8 lost jobs. At a $300 average ticket, roughly $2,250 a month, or about $27,000 a year, because of answers the owner already knew.

The fix costs an afternoon of writing.

Which Businesses Have the Easiest and Hardest Setups?

Easiest: businesses with a simple service menu and a digital calendar. Lawn care, cleaning, routine pest control, junk removal. Most calls are "what do you do, where do you go, when can you come."

Harder: businesses where pricing depends on an inspection, like roofing, foundation repair, or remodeling. These still work well. The AI's job is to book the inspection and collect details, and it should never quote the job.

Most complex: operations with dispatch rules across zones, crews, and job types. Expect more time on booking rules and calendar setup.

Extra steps: medical and dental offices, because of HIPAA and the BAA requirement.

Can't I Just Give It My Website?

This is the most common onboarding mistake. Owners send the website and say "it's all on there."

Usually it isn't. Websites tend to leave out the things callers actually ask. Do you go out to Tomball? Do you charge for estimates? Can you come today? Do you do commercial?

The opposite mistake is sending a 40-page policy binder that buries the five answers that matter.

Most blog posts on this topic also get the setup timeline wrong. Some say setup takes minutes: paste your site, pick a voice, done. You can get something talking in minutes, but you can't get something you trust with your customers that fast. Others imply it takes months and an IT team. For most small businesses, it's a few focused hours of the owner's time, then a couple of weeks of reviewing transcripts and tightening answers.

What's on the AI Receptionist Onboarding Checklist?

Here's what I collect, in this order.

1. Business basics. Name, hours, service area (cities and ZIP codes), services you offer, and services you don't.
2. Top 20 caller questions with your answers, written the way you'd say them on the phone.
3. Booking rules. Which calendar, appointment types, durations, buffers between jobs, and who takes which jobs.
4. Transfer rules. Who gets which calls, on which numbers, during which hours, and the fallback if nobody answers.
5. The "don't say" list. Topics the AI should hand off instead of answering.
6. Text messages. Booking confirmation, reminders, and the summary you want texted to you.
7. Compliance items. Recording disclosure, consent for texts, and a BAA for healthcare.
8. Testing. We call it ourselves, ask the hard questions, and fix what breaks.

What's One Thing to Do This Week?

Keep a notepad by the phone for five business days, or ask your office person to. Every time someone asks a question, write it down and add a tally mark when it repeats.

At the end of the week you'll have your top questions, ranked. Write your answer to each in one or two sentences, the way you'd actually say it.

Then write the "don't say" list. Most onboarding focuses on what the AI should answer, but what it shouldn't say matters just as much. Don't quote a firm price. Don't promise same-day service in July. Don't diagnose over the phone. Don't discuss an open insurance claim. Give each one a safe fallback, like "the owner will call you about that today."

That document is the most valuable thing you can bring to an AI receptionist setup, or to a free audit of how your phones run today. It's also how you'd train the next human you hire.

What Changed With AI Receptionist Setup?

Training a human receptionist used to take weeks of shadowing. Answering services worked from a script and mostly took messages. Early chatbots needed rigid decision trees, with every path mapped by hand.

AI receptionists now work from a plain-language knowledge base. You write answers the way you'd explain them to a new hire, and the AI handles the phrasing when a caller asks the same thing ten different ways.

The setup work moved from programming conversations to documenting your business. That's something any owner can do, and most already know every answer.

How long does AI receptionist setup take?

For most small businesses, a few focused hours to gather information, then a couple of weeks of transcript review to refine answers.

Does the AI need access to my scheduling software?

To book directly, it needs a calendar connection. Integration with field-service tools varies, so confirm yours is supported early in the process.

What if a caller asks something that's not in the knowledge base?

A good setup has a fallback: the AI says someone will follow up, captures the details, and texts you a summary.

Can I update the information later?

Yes, and you should. Read transcripts regularly and add answers whenever a new question shows up.

The Bottom Line

An AI receptionist is only as good as what you teach it. Bring your business basics, your top caller questions, booking and transfer rules, and a "don't say" list. Start with a week of tally marks by the phone, and you'll have most of it.

If you want to hear what a well-configured AI receptionist sounds like, try the live demo at foundonline.pro.