The Binding Quote Your Website Just Gave a Stranger

It is late, and a homeowner has a water heater leaking onto the garage floor. They find your site and type into the chat widget: "How much to replace it, and can you come tomorrow?"

The AI chatbot answers in a second. "A standard 50-gallon replacement runs around $1,800, and we can have someone out tomorrow morning."

To a stressed homeowner that is a relief. To your business it is a firm price and a firm date for licensed work, put in writing, by a system that has not seen the job, does not know your schedule, and has no idea whether the unit is gas or electric. The crew that shows up tomorrow inherits whatever the bot promised tonight.

The Bot Just Made Two Promises You Have to Keep

In two clauses, the chatbot did two things that carry weight. It committed to a price, and it committed to a timeline. Neither was an estimate from someone who looked at the work. Both were stated as fact, on your site, in your company’s name, in a transcript the homeowner keeps.

A trained estimator knows that a number is a negotiation and a date is a maybe until the job is seen. The bot does not know that. It was built to answer the question, so it answered, and it answered as if it had the authority to bind you. The rest of this is about why "we told it to give ranges" does not protect you when it does the opposite.

A Quote Is Not Small Talk

A firm statement about price or outcome is not chatter. Under the Uniform Commercial Code (§ 2-313), an affirmation of fact or a promise that becomes part of the deal creates an express warranty, and you do not need the words "warrant" or "guarantee" to make one. Tell a homeowner "this will fix the leak for good" or "it will be done Friday," and those are promises the business can be held to.

The Code covers the goods. The installation itself carries its own contract and its own common-law duty to do the work as promised. So a chatbot that reassures, quotes, and commits is drafting terms no estimator approved and no manager saw, and the homeowner is entitled to hold you to the friendliest version of every one of them.

In Most States, Bidding Itself Requires a License

Here is the part most owners miss. In most states you do not have to pick up a tool to break the licensing law. Advertising the work, offering to do the work, or submitting a bid without the right license is itself the violation. Washington calls unlicensed bidding a gross misdemeanor. California makes unlicensed contracting a misdemeanor with fines that climb on repeat. Nevada and Florida are written the same way, and a contract that grows out of unlicensed work can be void, with some states letting the homeowner recover everything they paid.

Now remember what a license actually is. It is permission for specific trades in a specific state. A bot does not check which. It will quote electrical work under a plumbing license, or a job one town over the state line, as cheerfully as it quotes anything else. Every quote it sends is a bid your business made, and the bot never once asked whether you were allowed to make it.

The Most Dangerous Answer Is the DIY One

Quotes cost money. This one can cost more than that. A homeowner asks, "my outlet is sparking, how do I fix it myself?" and the chatbot, built to be helpful, walks them through it. Or it explains how to relight a gas pilot, or how to get up on the roof to clear a valley.

If that homeowner is hurt, or burns down a kitchen, or floods a crawlspace following steps your website gave them, the business that handed out the instructions is in the path of the claim. There is a reason your techs are trained and your office tells callers to leave gas and wiring to a pro. A chatbot has no such instinct. It treats "how do I fix it myself" like any other question, and the most dangerous thing it can do is succeed at sounding like it knows.

The Follow-Up Text Has Its Own Rulebook

There is one more line that is easy to trip. If your site captures a phone number and your system answers with automated texts or calls, the Telephone Consumer Protection Act is in play. The statute restricts automated calls and texts to mobile phones without the consumer’s prior express consent, and it carries damages of $500 to $1,500 for every single message.

Courts in 2025 and 2026 have been trimming the FCC’s stricter consent rules, including the written-consent and one-to-one requirements, so the precise standard is shifting. What is not shifting is the shape of the lawsuit: a tool that harvests numbers and fires off automated outreach with no clean record of consent is the exact fact pattern these cases are built on.

The Disclaimer Does Not Cover the Crew

The usual answer is a line under the chat box: "estimates are not final," "subject to inspection," "for informational purposes only." It feels like cover.

It is not. A disclaimer does not turn a firm quote into an estimate after the homeowner has read the firm quote, and it does not turn unsafe instructions into safe ones. The sentence at the bottom describes what you hoped the bot would do. The conversation above it is what the bot actually did, and that is the part a homeowner, or a licensing board, will read back to you.

Why the Instruction Fails

Your vendor will tell you to fix this with a prompt. Tell the chatbot to give ranges, never firm numbers, and never tell anyone to do their own electrical work.

It will not hold, because the model only follows the rule when the question looks like the one it was warned about. The homeowner does not say "give me a firm bid." They say "ballpark, what am I looking at?" and the model gives a friendly, specific number because that is what sounds helpful. The instruction was loaded the whole time. It just did not recognize the sentence that bound you.

That is the gap. An instruction asks the model to behave. It does not stop the model from speaking. Architecture works the other way. A boundary built into the system decides what the agent may say before it answers, so a firm price, a committed date, or a do-it-yourself repair never leaves the widget, no matter how the question is phrased. "Will not" is a suggestion. "Cannot" is an architecture.

What It Costs When It Goes Wrong

The bill arrives from several directions. A binding quote you cannot meet is a breach you eat or a customer you lose. A licensing board can fine unlicensed bidding and bar you from collecting on the job. A homeowner hurt by DIY advice brings a negligence claim that your general liability carrier will read very closely. And an automated-outreach misstep carries per-message damages that multiply fast across a contact list.

Then there is the part no fine schedule lists. Home services runs on reviews and referrals. One bad transcript, screenshotted and posted, does more damage than the penalty attached to it.

The Question Before a Bot Quotes a Job

The chatbot on your site is not a dispatcher you can retrain after a bad call. It is a system that will quote a price, promise a date, and coach a homeowner through live wiring for anyone who asks, in writing, in your company’s name, at 11 at night when no estimator is awake.

So the question is not whether your AI is friendly. It is whether your AI can be made unable to quote a firm price, commit to a date, or give a repair instruction. If the answer is that you told it not to, you do not have a control. You have a promise the bot can break for you.

Fred is built the other way. It answers from your own content, captures the job details, scores the lead, and routes anything that touches price, scheduling, or a how-to repair to a real person on your team. It runs more than 50 industry guardrail packs, and the home services pack is built around the lines that bind a contractor. Fred does not hand out a firm quote and does not coach anyone through a gas line. It cannot. Fred answers what it should and books the rest for the people who carry the license.

Frequently asked questions