Fred: A Cleaning Assistant That Won’t Promise to Kill a Virus

A customer messages your cleaning site because someone in the house has been sick, and asks whether a deep clean will kill the virus, sanitize everything, and what it will cost. A general-purpose chatbot says yes on all counts and quotes a flat price. No one saw the home or checked what the crew’s products are actually approved to claim. The company wanted a tool to book more cleans. It got one making a public-health claim and a price commitment nobody authorized.

The threat and standard articles in this series walk through how that becomes a labeling and deceptive-practices problem. This one is the answer: what a compliant AI assistant for cleaning services looks like, and why the design, not a disclaimer, keeps a chat window from promising what the company cannot back.

The Real Choice Is Governed or Ungoverned

Cleaning companies ask whether to use AI on the site at all. Customers already decided. They want to ask about services, book a clean, and get a sense of pricing without waiting for business hours, and the company that makes them wait loses the job. The real question is whether the assistant is governed.

An ungoverned chatbot promises to kill viruses, quotes an unseen home, and vouches for the crew. A governed one books the clean, explains services in general, and routes health claims, firm prices, and trust questions to a person. Same convenience, completely different exposure when a customer quotes the transcript back.

How Fred Holds the Line

Fred starts from the opposite default of a general chatbot. A generic bot answers anything unless told not to. Fred answers only what it is cleared to, and the boundary lives in the system rather than a prompt that drifts when a customer pushes.

In practice, Fred works from your own content. It books cleans, explains your services and service area, and captures the details the crew needs. It does not promise to kill a virus or fully sanitize a home, because disinfectants are regulated as pesticides under FIFRA and the claims a service can make are tied to its products’ registered labels, with unsubstantiated efficacy claims also reaching the FTC’s deceptive-practices authority. It does not quote a home it cannot see, and it does not vouch for insurance or background checks unless that reflects your actual policy. Fred runs more than 50 industry guardrail packs, and the cleaning pack is built around disinfectant claims, blind quotes, and company representations.

The design proves itself under pressure. A worried customer will rephrase "will it kill the germs" several ways, and a prompt-instructed bot eventually answers the version it was not warned about. Fred does not rely on catching the phrasing. It decides what may be said before the reply forms, so the disinfectant claim never reaches the customer. "Will not" is a suggestion. "Cannot" is an architecture.

Fred vs. a Generic AI Chatbot

That single screen is the argument. A general chatbot is helpful until helpful becomes a regulated health claim or a blind quote. Fred is helpful across everything that moves a job forward and structurally incapable of claiming what the company cannot back.

What Your Company Actually Gets

Set the rules aside and look at the phone. Fred handles the after-hours questions that usually die in voicemail, do you clean my area, what does a deep clean include, how often do people book, can someone come this week, and answers them instantly, overnight, in plain language. It captures the lead with the context that makes the first call productive. Everything that touches a health claim, a firm price, or a trust representation stays with a person.

The captured-lead math is the real win. The customer who messages at 9 p.m. and gets a useful, accurate reply is the one who books, instead of the one who called the next company while waiting for a callback. Fred turns the after-hours window into a pipeline without ever making a claim it cannot defend.

So the question is not whether your competitors will put AI on their sites. They will. It is whether yours is governed before a customer holds it to a promise it never should have made.

Frequently asked questions