Most "AI chatbot" pitches get small-business owners exactly backwards. They sell you a personality — a friendly bot that "engages visitors." What a service business actually needs is not engagement. It is intake: catching the customer who shows up at 9pm with a broken garage door and turning them into a booked job before they call the next company on Google.
I will show you the difference using a real one I built and run — for a garage-door repair company in the Chicago suburbs.
The problem an intake assistant solves
Garage-door repair is an emergency trade. A spring snaps, a door will not close, the car is trapped inside — and that customer is buying tonight. But a small repair operation has two or three people, and they are under cars, not at a desk. Calls go to voicemail. After-hours leads evaporate. The website — usually a brochure with a phone number — does nothing while everyone is asleep.
That is the gap. Not "we need more visitors." This company's site already got plenty: search traffic was actually up year over year when the owner came to me convinced his business was getting worse. The visitors were arriving. They just were not turning into phone calls.
What the assistant actually is (under the hood)
It is not a generic chatbot reciting internet facts about garage doors. It is a retrieval-grounded intake worker:
- It answers from the company's own knowledge — their services, their pricing logic, their service areas — held in a vector database the assistant searches before it replies. So when someone asks "do you cover Palatine?" or "roughly what does a spring replacement run?", it answers like the company, not like Wikipedia.
- It runs on a current top-tier language model, but the model is the engine, not the product. The product is the wiring around it: it knows the business, and every conversation has one job — get this person to the next step (call now, or leave a number for an after-hours callback).
- It sits alongside the conversion plumbing that makes the rest of the site work: click-to-call links everywhere, structured data that tells Google exactly what this business does and where, an FAQ that answers the pre-call questions. The assistant is the part that works when no human can pick up.
What it does — and what it does not
Here is where I differ from the people selling AI chatbots: an intake assistant does not create demand. It stops you from leaking the demand you already have. This company did not have a traffic problem. It had a capture problem — visitors slipping away after hours, and (it turned out) some literally broken click-to-call buttons that meant the phone could not ring even when someone tried.
That is the whole philosophy. I would rather instrument the thing and show you a real before/after in 90 days than hand you a fake testimonial today. So that is exactly what we are doing on this build — measuring after-hours captured leads from day one, and I will publish the number whether it is good or not.
Should your business have one?
An AI intake assistant earns its place when three things are true: you get inquiries outside the hours you can answer them; each lost lead is worth real money (a service call, not a $5 sale); and your site already gets visitors who leave without converting. A garage-door company, a plumber, an HVAC shop, a clinic taking after-hours booking requests — yes. A business with five visitors a week — fix the traffic first.
If you are not sure which camp you are in, that is exactly what a quick site check tells you: whether your visitors are leaving without a way to reach you, and whether an intake layer would catch them.