Estimates, Express Warranties, and Your Contracting Site: A 2026 Playbook

The whole contracting model depends on one habit: quote from the actual job, never from a guess. Measurements, materials, permits, and whatever the demolition uncovers decide the real number, and a price stated before any of that is the seed of a dispute. An assistant on the website undermines that habit the instant it answers "what does this run" or "is it guaranteed," because a homeowner anchors on whatever the site said and the business gets held to it. The 2026 standard is about keeping the assistant useful for capturing the lead while leaving pricing, warranty, and scope with someone who has seen the work.

This guide is the companion to the threat piece. The threat side covers the number the website quoted and the crew was held to. This one covers the standard: what a compliant contracting assistant says, what it routes to a person, and why the line gets enforced in code rather than fine print.

A Number Anchors, Even When It Is Called a Ballpark

The hazard in a website quote is not imprecision; it is that the figure exists at all. A homeowner does not file away "around forty grand" as a caveat. They treat it as the price and feel misled when the real bid differs, and a contractor that quotes work in a way that misleads invites the kind of scrutiny the FTC applies to deceptive practices. A real estimate depends on inputs the website cannot see, so the compliant standard is that the assistant does not produce a number, ballpark or otherwise.

What it does instead is capture the project, the scope, the timeline, the homeowner’s goals, and book the site visit where a real estimate gets made. That is a useful, conversion-friendly job. It just stops short of putting a figure in a transcript the business never agreed to.

"Guaranteed" Is a Warranty the Business Defines

When an assistant calls the work "guaranteed" or "fully covered," it is not offering reassurance. It is stating a term. An affirmation of fact about the work can create an express warranty under the Uniform Commercial Code’s warranty rules where goods are involved, with parallel principles in service contracts. The contractor decides what it warrants and for how long, and a sentence in chat should not be allowed to rewrite those terms before a contract even exists.

So the standard treats warranty language as contract language. The assistant points to the company’s actual warranty and routes any "is it covered" question to a person, rather than improvising a promise the business will have to honor.

What AI Compliance for Contractors Requires Around Licensing and Scope

Most states license contractors and restrict who may bid on, perform, or advise on certain work. An assistant that cheerfully says "yes, we can handle that" for a trade outside the company’s license, or that drifts toward design or engineering guidance, can put the business on the wrong side of those rules. A website has no way to judge scope or qualifications, so it should not try.

The compliant standard pulls these threads into a short list. The assistant captures the project and books the visit; it does not quote a price. It points to the company’s real warranty; it does not promise coverage. It describes the services the company actually offers; it does not commit to scopes or trades it cannot judge. And it keeps a record of what it told the homeowner, so a later dispute over the bid or the warranty has a clean trail rather than a freelancing transcript.

Why a Prompt Cannot Meet the Standard

The reflex is to write the rules into the assistant’s instructions: never quote, never say "guaranteed," never commit to scope. Treat the boundary as set.

It is not set, because of how the model works. It follows an instruction when the request matches the wording it was warned about, and a homeowner shopping three bids does not use that wording. You tell it never to quote. The homeowner does not ask for a quote. They lean in, "just ballpark it, are we talking thirty grand or fifty?" The model reads a reasonable person who wants a range and gives one, because being helpful under pressure is its default and a prompt is only a request to hold that default back. The rule was loaded the whole time. It just never recognized the sentence that crossed the line.

That is the gap between an instruction and a standard. An instruction asks the model to behave. It does not stop the model from speaking. A real boundary is enforced in the system and decides what the assistant may say before it answers, so a price, a warranty promise, or a scope commitment never reaches the homeowner no matter how the question is phrased. "Will not" is a suggestion. "Cannot" is an architecture.

What a Compliant Deployment Looks Like

Meeting the standard does not mean a contractor gives up the assistant that captures leads off the project pages overnight. It means running one built to keep pricing, warranty, and scope with a person who has seen the job.

Fred is built that way. It answers from your own company content, captures the project, books the site visit, and routes pricing, warranty, and scope questions to your team. It runs more than 50 industry guardrail packs, and the contracting pack is built around express-warranty exposure, deceptive-quoting risk, and the licensing-scope line. Fred does not throw out a number or call the work guaranteed. It cannot. It handles the intake, books the visit, logs every exchange, and gets the regulated commitments to a person.

The aim is not a quicker quote. It is a website that cannot price a job it never saw or warrant work the company never agreed to cover.

Frequently asked questions