AI Receptionist
Giving estimates by phone: what an AI receptionist can and can’t quote
“How much will this cost?” is the first question most callers ask, and it is the question phone systems handle worst. A human dispatcher who has worked the trade for years can hear “water heater” and give a range that is usually right. An AI receptionist has no instincts. It has whatever pricing rules you gave it. That difference is the whole story of estimates by phone: the system is only as good as the guardrails around it.
The fear most owners have is simple. They picture the AI confidently quoting $400 for a job that costs $1,200, and the tech arriving to an angry customer holding a recorded promise. That scenario is real if the system is allowed to guess. It is entirely preventable if the system is allowed to quote only what you told it to quote.
What an AI receptionist can quote safely
There is a whole category of pricing that causes no trouble on the phone: fixed, published numbers. The diagnostic or service call fee. The trip charge. Standard maintenance visit prices. Membership plan prices. These are the same numbers a dispatcher reads off a card, and an AI reads them off the same card with the same accuracy. If your shop has a price book for common flat-rate jobs, those numbers can go on the card too.
The rule that makes this safe is boring and absolute: the system quotes only numbers from the approved list, and everything else gets a range or a handoff. “Our diagnostic visit is $89, and the tech gives you the full price before any work starts” is a complete and honest answer. It tells the caller what they will pay today and who decides the rest.
Where the line is: ranges, not promises
Most real jobs cannot be priced without seeing them. A water heater replacement depends on the unit, the venting, the code upgrades. An electrical panel issue could be a breaker or a rewire. For these, the honest phone answer is a range with the variables named: “Most water heater replacements we do run between X and Y, depending on the unit size and venting. The tech confirms the exact price on site before starting.”
The key detail is that the range comes from your history, not from the AI’s imagination. You supply the ranges. The system repeats them. If a job type has no approved range, the system does not invent one. It books the diagnostic visit and sets the expectation that pricing happens on site. That is not a limitation of the technology. It is the policy working as designed.
The handoff moments worth scripting
Some calls need a human estimator no matter what. Large projects, anything involving permits or structural work, commercial jobs with bid requirements, and callers who are clearly comparing detailed quotes all deserve a person. The AI’s job in these moments is not to quote. It is to qualify and route: capture what the project is, where it is, and when the caller is available, then get it to the right person with notes.
A related moment is the price shopper. “I just need a ballpark” usually means “I am calling five shops and picking the cheapest number.” The honest play is the diagnostic fee plus the range, and an offer to book the visit. Shops that chase price shoppers with made-up numbers win the jobs they will regret. Our playbook for turning price shoppers into booked jobs goes deeper on this conversation.
Setting up the guardrails
The setup is a pricing document, not a software project. List the numbers the system may say: diagnostic fee, trip charge, flat-rate services, membership prices. List the ranges it may give, with the variables that move them. Then write the sentences for everything else: “That one needs eyes on it before we can price it. I can get a tech out [window] for the $89 diagnostic, and you will have the full price before any work starts.”
Review the document quarterly, because prices move and the phone script should move with them. And listen to a few estimate calls in the first month. You are listening for one thing: did the system stay inside the lines. If it did, the guardrails work. If it strayed, the document needs another line, not a new system.
For the dispatch side of the same operation, see AI receptionist for plumbers: dispatch, quotes, and after-hours, and for the follow-up that turns quotes into revenue, quote follow-ups that win the job back. If you are still deciding whether callers will accept an AI on estimates at all, the honest answer on whether customers know it is AI is worth a read.



