Will an AI Receptionist Answer Customer Questions About My Services Correctly?

It's a Saturday morning in The Woodlands. A homeowner walks the yard and sees two brown patches where the sprinklers clearly aren't reaching. She calls a lawn and irrigation company and asks three questions in one breath: "Do y'all fix sprinkler heads, do you work on Rainbird controllers, and is there a charge just to come out?"

If the company's AI receptionist has good answers to those three questions, she books a visit. If it says "someone will get back to you," she calls the next company. If it makes up a trip charge, the owner has a problem on Monday.

All three outcomes come down to what the owner wrote in the knowledge base.

The short answer

An AI receptionist's knowledge base is the set of approved answers it uses on calls: your services, service area, hours, policies, and any price ranges you're comfortable sharing. Modern voice agents understand questions asked many different ways and match them to those answers. Accuracy depends on how complete and clear the knowledge base is, and on a fallback rule for questions it doesn't cover.

What does the research say about answering customer questions?

Callers with an unanswered question don't wait long.

• Calls go unanswered constantly. Small-business call studies put unanswered inbound calls somewhere between roughly 40% and over 60%, depending on the study and industry.
• Most callers don't try twice. A commonly cited industry figure says about 85% of people who don't get through won't call back.
• Answers are moving into AI everywhere. Gartner predicted in February 2024 that traditional search engine volume would drop 25% by 2026 as people turn to AI chatbots and virtual agents. Pew Research Center found in 2025 that Google users clicked a traditional result on about 8% of visits when an AI summary appeared, compared with about 15% without one.

That last point matters for more than your phone. The same clear question-and-answer content that makes an AI receptionist accurate is the kind of content AI search tools pull from when they answer questions about local businesses. That's the core of answer engine optimization. Writing it once helps on both fronts.

What does it cost when questions go unanswered?

Here's math you can redo with your own numbers.

Say an irrigation company gets 30 calls a week, and 12 of them are pre-booking questions like "do you do X" or "what does Y cost." If nobody answers 40% of those, that's about 5 callers a week with an unanswered question. If 2 of them would have booked a $200 repair, that's $400 a week, or about $20,800 a year.

The cost of a wrong answer is different. Quote a trip charge you don't have, or promise a service you don't offer, and you'll spend time on damage control and maybe lose a review. That's why the fallback rule matters as much as the answers.

Which businesses can rely on AI to answer service questions?

Strong fit: the same questions every day. Lawn and irrigation, pest control, pool service, HVAC maintenance plans, house cleaning, junk removal, med spas. Answers are stable and a good knowledge base covers most of them.

Good fit with limits: inspection-based pricing. Foundation repair, roofing damage, mold, custom remodels. The AI can explain your process, what the inspection involves, what affects the price, and then book the inspection. It shouldn't guess a number.

Healthcare: general service and scheduling questions work, but call data is protected health information. HIPAA applies, so use a vendor that signs a BAA.

Will the AI make things up?

This is the fear I hear most, and it's fair. General-purpose chatbots will sometimes answer confidently with something wrong.

A well-scoped AI receptionist is a different setup. It works from an approved knowledge base, and it has explicit instructions to hand off anything outside that knowledge base. When an answer comes out wrong in testing, the first place I look is a missing or vague entry. The fix is usually writing a better one. (If a wrong answer does reach a customer, here's what to do when an AI receptionist gives wrong information.)

The opposite myth is that AI "knows" your business once it's connected. It knows only what you give it. It doesn't know you stopped serving Conroe last spring unless you told it.

The most common mistake is pasting your website into the AI and calling it done. Websites are written to sell, and they skip what callers actually ask: "Do you work Saturdays?" "Is there a trip charge to Magnolia?" "Do you service Hunter controllers?"

The second mistake is leaving pricing out completely, so every price question becomes "someone will call you." If you're comfortable with ranges, share them.

I learned the wording lesson on my own line. FoundOnline's services have names that make sense to me: AEO, GEO, database reactivation. A caller doesn't ask for "GEO." They ask, "Can you help me show up when people search for a roofer?" Writing knowledge base entries as the questions people really ask, in their words, mattered more than anything technical. We're early and don't publish client results yet, but that applies to any business.

How do you build a knowledge base for an AI receptionist?

This is the sequence I follow.

1. Collect real questions from call notes, texts, emails, Google Business Profile questions, and the owner's memory.
2. Write each answer the way you'd say it on the phone. Short, specific, and in plain words.
3. Set boundaries. Price ranges you'll share, services you don't offer, areas you don't serve.
4. Write the fallback. Exactly what the AI says and does when a question isn't covered.
5. Test tricky phrasings. Ask the same question five different ways, with slang and loose phrasing like "y'all do sprinklers?"
6. Launch and review transcripts weekly. Every gap becomes a new entry.
7. Set a calendar reminder to update answers when prices, hours, or services change.

What's one thing I can do this week?

Open your texts and email from the last month and copy every question a customer asked into one document. Group the duplicates. Write your answer under each one, the way you'd say it on the phone.

Mark which answers include a price and decide whether you're comfortable sharing a range.

That document pulls triple duty. It's a website FAQ page, a training doc for your next hire, and the core of an AI receptionist's knowledge base.

Then write one more answer on purpose: the "I don't know" answer. Something like, "That's a good one for the owner. Let me grab a couple of details and have him call you back today." The AI collects the details and texts you. Read the transcripts each week and turn every "I don't know" into a new entry. The knowledge base grows from real calls.

What changed with AI?

Phone trees could route a call but couldn't answer anything. Early chatbots worked from rigid scripts and broke the moment a customer phrased something differently. Answering services had a short script and a message pad.

Today's voice agents run on large language models that understand questions however they're asked. "Y'all do sprinklers?" and "Do you repair irrigation systems?" land on the same answer. They pull from your knowledge base instead of a fixed decision tree.

That shifted the work. You used to need someone to program every path. Now the job is writing clear, accurate answers about your own business. Nobody knows those answers better than you do.

Can an AI receptionist give prices?

It can share the price ranges you approve. For work that needs an inspection, it should explain what affects the price and book the visit.

What happens if a caller asks something it doesn't know?

With a good setup, it says so, collects the caller's details, and hands off to you, usually with a text summary. Write that fallback before launch.

How often should I update the knowledge base?

Whenever prices, hours, services, or service areas change, and after each weekly transcript review in the first month.

Can I just upload my website?

It's a starting point, but websites skip most real caller questions. Add a question-and-answer section built from actual calls.

The bottom line

An AI receptionist answers service questions as well as its knowledge base is written. Use your customers' words, approve your price ranges, and give it a clear "I don't know" path. Then let real calls tell you what to add.

If you want to hear how an AI receptionist handles questions it does and doesn't know, try the live demo on foundonline.pro.