What should go in an AI receptionist's knowledge base?

Start with the questions callers ask most. Hours, including after-hours and holiday rules. Services you offer and the ones you do not. Prices you are willing to quote, like a service call fee or a starting price. Your service area, by town or by miles. Policies on estimates, deposits, and payment. What counts as an emergency. Finally, who to transfer to and when. Two or three short sentences per item is plenty.

Do I need to know how to code to build a knowledge base?

No. You write it the way you would explain your business to a new hire on their first day. Plain sentences, one fact each. Most AI receptionist services give you a form or a simple editor for this. Cira includes white-glove onboarding at no extra cost, so a person helps you fill it in. Setup takes minutes, not days, and you can edit it any time.

What happens when a caller asks something that is not in the knowledge base?

A good AI receptionist says it does not have that answer, then takes a message or transfers the call to a person. It should not guess. If you see the same gap in a few call summaries, add the answer to the knowledge base and the AI handles it from then on. Reading your own call summaries is the fastest way to find the gaps.

Glossary

AI Tech

What Is an AI Knowledge Base?

By Bryan Smith, CEOUpdated August 20, 2026

In an answering service, a knowledge base is the set of business facts an AI receptionist answers from: hours, prices, service area, services offered, and policies. The owner writes it in plain language. The AI uses it on every call, so callers get the same correct answers a well-trained front desk person would give.

Key Takeaways

The knowledge base is what the AI knows about your business. It is not the AI itself.
It is written in plain sentences by the owner, not in code.
The most useful entries are the awkward ones: what you do not do, what you charge to show up, and where you will not drive.
A stale knowledge base gives confident wrong answers. Update it when prices or hours change.

How an AI Knowledge Base Works

An AI receptionist has two parts. The brain is a large language model that works out what the caller means and forms a reply. (A separate speech tool turns the caller's voice into text first.) The memory is the knowledge base. The brain is the same for every business. The knowledge base is what makes it yours.

On a call:

You fill in the facts. Hours, services, prices, service area, policies, emergencies. Plain sentences, like you are training a new hire.
A caller asks a question. "Do you come out to Cheney?" or "What does a service call cost?"
The AI finds the matching fact. It does not read a script. It finds the answer and says it in natural words.
If the fact is not there, it says so. A well-built AI voice agent admits it does not know, then takes a message or transfers.
You update it as things change. New price, new town. Edit the entry and the next call uses it.

The same knowledge base can serve phone, text, and website chat at once, so every channel gives the same answer.

Example of an AI Knowledge Base

A plumbing company with a crew of three gets about 200 calls a month. A few entries from the owner's knowledge base:

Hours: 7 am to 6 pm, Monday through Friday. After hours, take a message unless it is an active leak, then transfer to the on-call tech.
Service call fee: $89, applied to the job if they book the repair.
Service area: within 30 miles of the shop. Cheney yes, Ritzville no.
Water heater replacement starts at $1,400. Do not quote repipes over the phone. Book a free estimate instead.

At 8 pm a homeowner calls: "Do you service Cheney, how much to come out, and can you be here tomorrow?" The AI answers all three from the list, books the 8 am slot, and sends a summary. The job comes to $400.

Now the math on time. If about 1 in 4 of those 200 calls is a simple hours, price, or area question, that is roughly 50 calls a month. At 3 minutes each, the owner gets back about 150 minutes, or two and a half hours, every month. That is before counting the jobs booked at night that would have gone to voicemail.

What People Get Wrong About AI Knowledge Bases

Owners write the knowledge base like the About page on their website. Founded in 2009. Family owned. Proud of our work. The AI does not need any of that. Callers never ask about it.

What callers ask is sharper and a little awkward. Do you charge to come out? Do you do mobile homes? Do you go to my town? Do you work Saturdays? Is the estimate free? Those are the questions that decide whether the job gets booked, and they are the ones most owners leave out, because the answer is sometimes "no" or "it depends." So the AI either guesses or says it does not know, on the exact question that mattered.

The fix is to build the knowledge base from real calls, not from marketing. Pull your last 30 call summaries or transcripts. Write down every question a caller asked. Answer each in one or two plain sentences, including the "no" answers. That list is your knowledge base. Reread it every time a price or a rule changes, because a confident wrong answer is worse than no answer.

AI Knowledge Base vs. Call Script vs. Help Desk Software

A call script is the exact words a human operator reads. A knowledge base is facts the AI turns into its own words, so it can handle questions the script never planned for.
A help desk knowledge base is a public library of how-to articles that customers search themselves. The AI kind is private, short, and built for answering live questions in seconds.
A CRM stores facts about your customers: names, addresses, past jobs. The knowledge base stores facts about your business. The AI uses both, but they are different lists.

Why It Matters

The knowledge base is the difference between an AI that takes messages and an AI that books jobs. With no facts, the best brain in the world can only say "someone will call you back." With twenty good entries, it quotes the fee and fills the calendar. Setting up an AI receptionist is mostly this step. Forwarding the phone takes two minutes. The knowledge base is where the owner's time goes, and it pays back every month. Our guide on how to set up an AI receptionist walks through it entry by entry.

The Bottom Line

An AI knowledge base is the list of plain-language facts an AI receptionist answers from: hours, prices, service area, services, and policies. It is the part that makes a generic AI yours. Build it from the questions real callers ask, include the awkward "no" answers, and update it whenever the business changes. Do that, and the AI books jobs instead of taking messages.

Frequently Asked Questions

What should go in an AI receptionist's knowledge base?: Start with the questions callers ask most. Hours, including after-hours and holiday rules. Services you offer and the ones you do not. Prices you are willing to quote, like a service call fee or a starting price. Your service area, by town or by miles. Policies on estimates, deposits, and payment. What counts as an emergency. Finally, who to transfer to and when. Two or three short sentences per item is plenty.
Do I need to know how to code to build a knowledge base?: No. You write it the way you would explain your business to a new hire on their first day. Plain sentences, one fact each. Most AI receptionist services give you a form or a simple editor for this. Cira includes white-glove onboarding at no extra cost, so a person helps you fill it in. Setup takes minutes, not days, and you can edit it any time.
What happens when a caller asks something that is not in the knowledge base?: A good AI receptionist says it does not have that answer, then takes a message or transfers the call to a person. It should not guess. If you see the same gap in a few call summaries, add the answer to the knowledge base and the AI handles it from then on. Reading your own call summaries is the fastest way to find the gaps.

Never miss another call

Cira answers every call, books jobs, and texts you the details while you work.

Start Free Trial Back to Glossary