AI Insights Professional & B2B
Fred: A Firm Assistant That Won’t Render an Opinion It Can’t Seal
Talk to Fred
Ask Fred about Professional & B2B
This is the same Fred you would put on your own site. Ask about Professional & B2B, compliance, or how the guardrails work. Fred listens.
A prospective client messages your firm’s site about removing a wall and asks whether it is load-bearing and whether an open plan would meet code for their house. A general-purpose chatbot says the wall is probably fine to remove and the plan should meet code. No one looked at the structure or pulled the local code. The firm wanted a tool to qualify leads. It got one issuing professional opinions that licensure reserves for an architect or engineer who has examined the actual project.
The threat and standard articles in this series walk through how that becomes unlicensed practice and a dangerous renovation. This one is the answer: what a compliant AI assistant for architects and engineers looks like, and why the design, not a disclaimer, keeps a chat window from rendering an opinion no one can seal.
The Real Choice Is Governed or Ungoverned
Firms ask whether to use AI on the site at all. Clients already decided. They want to describe a project, ask about the firm’s work, and book a consultation without waiting for business hours, and the firm that makes them wait loses the project. The real question is whether the assistant is governed.
An ungoverned chatbot calls a wall non-structural, says a plan meets code, and quotes a fee. A governed one books the consultation, explains the firm’s process, and routes structural calls, code opinions, and fees to a licensed professional. Same convenience, completely different exposure when a client renovates on the strength of the website.
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 client pushes.
In practice, Fred works from your own content. It explains your services and process, qualifies the project, and captures the inquiry. It does not say whether a wall is load-bearing or whether a design meets code, because determining that is the practice of architecture or engineering, the very work the professional seal represents and that licensure reserves for a licensed person who has examined the project. It does not render feasibility verdicts, and it does not commit a fee or a timeline, because a promise the firm cannot keep can draw the FTC’s deceptive-practices authority. Fred runs more than 50 industry guardrail packs, and the architecture and engineering pack is built around the practice-of-licensure line, the seal, and fee commitments.
The design proves itself under pressure. A client who wants an answer will rephrase "is the wall structural and will this pass" 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 structural call never reaches the client. "Will not" is a suggestion. "Cannot" is an architecture.
Fred vs. a Generic AI Chatbot
| Situation | Generic AI Chatbot | Fred |
|---|---|---|
| "Is this wall load-bearing?" | Renders a structural opinion | Books a consultation with a licensed pro |
| "Will this meet code?" | Gives a code verdict | Routes code questions to a professional |
| "Will the city approve it?" | Offers a feasibility judgment | Captures the project; defers to the firm |
| "What will it cost?" | Quotes a fee and timeline | Routes fees to a proposal a person prepares |
| Where the rules live | In a prompt a client can push past | Built into the system; enforced before output |
| Who owns the regulated answer | Effectively the bot, and your firm | The licensed professional who examines the project |
That single screen is the argument. A general chatbot is helpful until helpful becomes a structural opinion or a code verdict. Fred is helpful across everything that moves a project forward and structurally incapable of rendering an opinion that belongs under a seal.
What Your Firm Actually Gets
Set the rules aside and look at the inbox. Fred handles the after-hours questions that usually sit until morning, do you handle projects like mine, how does your process work, what are the next steps, how do I set up a consultation, and answers them instantly, overnight, in plain language. It qualifies the project so the consultation starts warm. Everything that touches a structural call, a code opinion, feasibility, or a fee stays with a licensed professional.
The captured-lead math is the real win. The client 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 firm while waiting for a callback. Fred turns the after-hours window into a pipeline without ever rendering an opinion it cannot stand behind.
So the question is not whether your competitors will put AI on their sites. They will. It is whether yours is governed before a client renovates on an opinion it never should have given.
Frequently asked questions
Can a compliant assistant answer any design questions?
It can explain the firm’s services, process, and the kinds of projects you take on, the same context a good intake gives. What it will not do is say whether a wall is structural or whether a design meets code, because those are professional opinions licensure reserves for a licensed architect or engineer who has examined the project. Fred books the consultation and routes those questions to a professional.
Why won't it give a quick fee or timeline?
Because a quoted fee or timeline reads as a commitment clients plan around, and the terms depend on a scope only a person can set. A promise the firm cannot keep can also raise deceptive-practices concerns. Fred captures the project and routes fees and timelines to a proposal a person prepares.
How is Fred different from a generic chatbot told not to give opinions?
A prompt instruction is something the model abandons when a client rephrases "is the wall structural and will this pass" in a way it did not anticipate. Fred enforces its boundaries at the system level, deciding what is allowed before it responds, so a structural opinion or a code verdict is never generated regardless of wording. That is the difference between a bot that usually deflects and one that cannot render the opinion.