Should loan officers use an AI receptionist? What it can do, and what it must never say

Note: I’m an editorial persona for our compliance desk, not a lawyer. This explains public rules so you can plan with your compliance team. It isn’t legal advice.

A newly licensed loan officer usually has no assistant, a phone that rings during closings, and borrowers who call at 8:45 on a Sunday night. An AI receptionist is an obvious fix. It’s also new enough that many LOs — and their compliance teams — aren’t sure what it’s allowed to say.

The answer is a lot less than a loan officer, and that’s the point.

How much of the week is nobody answering?

Take a fairly generous office schedule — the hours our own launch desk keeps: 44 hours a week. That leaves the rest of the week’s 168 hours with nobody on the line.

Borrowers house-hunt evenings and weekends. Agents call between showings. A missed call often becomes a call to someone else.

What does an AI receptionist actually do?

A good one does five things, in this order:

1. Answers honestly. “Thanks for calling Jordan’s office — this is Jordan’s AI assistant, and this call may be recorded.”
2. Understands the reason. Buying, refinancing, a document question, an agent checking on a file.
3. Answers general questions only — from answers you’ve approved. Programs you offer, what to bring, how the process works, your hours.
4. Books or transfers. It offers times from your calendar, or transfers live to you if you’ve set rules for that.
5. Reports. Caller details, the reason, the booking and a transcript go to your CRM with an alert.

What must an AI receptionist never say?

This is where mortgage differs from a dentist’s office. The assistant should be configured so it never:

• Quotes rates, APRs, fees, closing costs or payment amounts
• Says a caller qualifies, is pre-approved or “should be fine”
• Gives individual financial, tax or legal advice
• Asks for Social Security numbers, account numbers or income documents
• Treats callers differently based on who they appear to be
• Keeps going after a caller asks for a person

Why so strict? Rates and payments in a marketing context can trigger disclosure rules under Regulation Z §1026.24, and an off-the-cuff “you’d probably qualify” can become a borrower complaint. Fair lending law (including the Equal Credit Opportunity Act and Regulation B) means every caller should get the same information and the same path to you.

The legal points to check with your compliance team

Inbound vs outbound

Answering calls people make to you is different from calling people. In February 2024, the FCC issued a declaratory ruling that AI-generated voices are “artificial” voices under the Telephone Consumer Protection Act. That means outbound calls using an AI voice need the consent the TCPA requires for artificial or prerecorded voice calls. If you plan to have an AI call leads back, get your consent language right first — or keep AI to inbound answering and texting where consent exists.

Recording notices

Call recording laws vary. Some states, California among them, require all parties’ consent to record a confidential communication. The practical standard: a recording notice in the first sentence of every call, and an option to turn recording off.

Disclosure that it’s an AI

Say so up front. Callers who find out later feel tricked; callers who are told usually don’t mind, especially when the AI gets them booked quickly.

Quiet hours for callbacks

If the assistant (or you) makes marketing calls or texts, the TCPA rules restrict telephone solicitations before 8 a.m. and after 9 p.m. in the recipient’s local time (47 CFR 64.1200).

Is an AI receptionist worth it for a new loan officer?

Our AI Receptionist is $149/mo with 300 minutes included, then $0.19 a minute. Paired with a starter website it’s the Day-One Desk bundle at $199/mo.

How to set one up well

A sample opening line

“Thanks for calling Jordan Example’s office. I’m Jordan’s AI assistant and this call may be recorded. I can answer general questions and book a time with Jordan — how can I help?”

A sample hard stop

“That’s a great question for Jordan, because it depends on your situation. I can’t quote rates, but I can book you a call. Would tomorrow at 5:30 work?”

Building the approved answer library

The assistant is only as good — and as compliant — as the answers you give it. Start with twenty questions you actually get, and write the answer your compliance team would be happy to hear on a recording.

Notice the pattern: answer what’s general, decline what’s specific, and always offer the next step.

What compliance teams usually ask — and good answers

“How do we know it won’t quote rates?” Rates, fees and approvals are configured as hard-stop topics with a scripted handoff, and you review transcripts in the first weeks to confirm.

“Is it recording?” Recording is a setting. If it’s on, the notice is the first sentence of every call.

“Will it treat callers differently?” No — it uses the same approved answers and the same booking path for everyone.

“Who sees the transcripts?” The loan officer and anyone you grant access to in the CRM. Keep the retention period aligned with your company’s policy.

“Can it call people?” Keep it inbound unless consent for AI-voice outbound calls is in place, given the FCC’s 2024 ruling.

What a sensible setup timeline looks like

• Days 1–2: collect hours, calendar, programs, states and the twenty most common questions.
• Days 3–4: write the answer library and hard stops; compliance reviews them.
• Day 5: test calls — including deliberately tricky ones — and adjust.
• Week 2: go live after hours first; review every transcript.
• Week 3 onward: expand to busy-time overflow once you trust the answers.

Starting after-hours only is a low-risk way to see how callers respond before the assistant ever answers during your working day.

What to watch in the first month

Read every transcript in week one. You’re looking for three things: questions it couldn’t answer (add approved answers), places it was too wordy (shorten), and any moment it drifted toward numbers (tighten the guardrail). After a month, weekly spot checks are usually enough.

An AI receptionist won’t close a single loan. What it does is make sure the person who wanted to talk to you actually gets on your calendar — and that every one of them hears the same careful, compliant answer.

Written by

Rosa Quintero, Compliance & AI Editor

Rosa writes about mortgage advertising rules, consent, and how AI receptionists and chat should behave for loan officers.

Editorial persona — GetsMortgageNow compliance desk · Not a lawyer; cites primary sources (CFPB, FCC, HUD)