AI Insights Financial Services
Fred: A Borrower Assistant That Won’t Quote Rates or Trip Reg B
Talk to Fred
Ask Fred about Financial Services
This is the same Fred you would put on your own site. Ask about Financial Services, compliance, or how the guardrails work. Fred listens.
A prospective borrower visits your lending site after hours and asks the question that drives every mortgage inquiry: "What rate can I get, and would I even qualify?" A general-purpose chatbot answers. It quotes a rate, offers an encouraging read on approval odds, and does it under your company’s name and NMLS number. The lender wanted a tool to capture leads while loan officers slept. It got one making advertised-rate and prequalification representations that land in the middle of Reg Z, Reg B, and fair lending.
The threat and standard pieces in this series explain how that exposure stacks up. This article is the answer: what a compliant AI assistant for mortgage lenders looks like, and why the design is what keeps the site clear of the advertising, eligibility, and fair-lending rules at once.
The Real Choice Is Governed or Ungoverned
Lenders ask whether to use AI on the site at all. Borrowers settled it, they expect instant answers about the process, documents, and next steps, and the lender that makes them wait loses them to one that did not. The decision that matters is whether the assistant is governed.
An ungoverned chatbot quotes a rate, hints at approval, and shapes what it tells a borrower based on what it infers about them. A governed one answers the process questions and routes every rate, prequalification, and eligibility question to a licensed loan officer. Same responsiveness, completely different posture under the lending rules.
How Fred Holds the Line
Fred runs on the opposite default from a general chatbot. A generic bot answers everything unless told not to; Fred answers only what it is cleared to, with the boundary enforced by the system rather than a prompt a determined borrower can talk around.
Concretely, Fred works from your own content. It explains how the loan process works, what documents to gather, the difference between prequalification and preapproval in general terms, and it captures the borrower with context. It does not quote a rate, because a number on your site is an advertised term governed by Reg Z’s advertising rules and triggers disclosure obligations the website never satisfies. It does not tell anyone whether they qualify, because eligibility runs through Reg B / ECOA and the Fair Housing Act’s bar on discrimination in lending, and a casual "you probably won’t qualify" is the exact kind of statement that invites a fair-lending complaint. It also stays out of work reserved for a licensed originator under the SAFE Act. Fred runs more than 50 industry guardrail packs, and the mortgage pack is built around the line between explaining the process and quoting a loan.
The strength shows under pressure. A borrower fishing for a number will rephrase "what rate" several ways, and a prompt-instructed bot eventually answers the version it was not warned about. Fred does not depend on recognizing the phrasing. It decides what may be said before the answer forms, so the rate quote never reaches the borrower.
Fred vs. a Generic AI Chatbot
| Situation | Generic AI Chatbot | Fred |
|---|---|---|
| "What rate can I get?" | Quotes a rate, an advertised term | Explains the process; routes the rate to a loan officer |
| "Would I qualify?" | Hints at approval odds | Routes eligibility to a licensed originator |
| Tailoring answers to the person | May vary by inferred traits, risking fair lending | Treats every borrower’s questions the same |
| Advertised loan terms | States them without Reg Z disclosures | Never advertises a term the site can’t disclose |
| Where the rules live | In a prompt the borrower can talk around | Built into the system; enforced before output |
| Who owns the regulated answer | Effectively the chatbot, and your NMLS | A licensed loan officer, every time |
That single screen is the argument. A generic tool is helpful until helpful becomes a rate quote or an eligibility call. Fred is helpful everywhere that carries no lending risk and structurally silent everywhere that does.
What Your Pipeline Actually Gets
Set the rules aside and look at the after-hours funnel. Fred answers the questions that otherwise sit in an inbox until morning, how the process works, what documents to prepare, how long things take, what the next step is, and it answers instantly, overnight, in plain language. It captures the borrower with context, so the loan officer who follows up already knows the goal and the timeline. Everything that quotes a rate, judges eligibility, or advertises a term stays with a licensed human.
The fair-lending angle is the quiet protection. A consistent, governed assistant gives every borrower the same accurate process information, which is exactly the opposite of the inconsistent, inferred-trait responses that draw fair-lending scrutiny. Instead of a liability sitting on the homepage, the after-hours window becomes a clean, even-handed top of funnel.
So the question is not whether your competitors will run AI on their sites. They already are. It is whether yours is governed before a borrower asks it the one thing it must never answer.
Frequently asked questions
Can a compliant assistant talk about rates or qualification at all?
It can explain how rates generally work, what affects them, and the difference between prequalification and preapproval in plain terms, and it routes any actual number or eligibility question to a licensed loan officer. What it will not do is quote a specific rate, which is an advertised term under Reg Z, or tell a borrower whether they qualify, which runs through Reg B/ECOA and fair lending. Fred explains the process and leaves the regulated answers to a licensed originator.
How does this protect us on fair lending?
Fair-lending risk grows when borrowers get inconsistent answers or treatment that varies by inferred characteristics. Because Fred gives every borrower the same governed, process-level information and never improvises eligibility calls, it removes a common source of that inconsistency. The Fair Housing Act and ECOA bar discrimination in lending; an assistant that cannot make eligibility judgments cannot make a discriminatory one.
How is Fred different from a generic chatbot told not to quote rates?
A prompt instruction is something the model drifts from when a borrower rephrases "what rate" in a way it did not anticipate. Fred enforces its boundaries at the system level, deciding what is allowed before it responds, so a rate quote or an approval hint is never generated regardless of wording. That is the difference between a bot that usually deflects and one that cannot give the answer.