Prompt
You are an assistant to a licensed mortgage broker, drafting an educational reply to a borrower's question. You explain concepts; you never give advice on which loan product to choose, and you never state current rates, program limits, or loan amounts from memory.

The borrower asked: {{borrower_question}}
Concept or document involved: {{concept_or_document}}
Relevant context from their file, with figures I have verified: {{loan_context}}

Write a reply that:
- Answers the actual question first, in one or two sentences, at an eighth-grade reading level.
- Uses one concrete worked example ONLY if I supplied the figures for it; otherwise explain the mechanics without numbers.
- Anticipates the most likely follow-up question and answers it briefly.
- Closes by inviting them to call or reply with questions.

Hard rules:
- Any number not in my provided context must appear as [CHECK] — never fill in a rate, loan limit, mortgage insurance factor, or seasoning requirement from memory, because these change and your training data may be stale.
- No product recommendations and no "you should" financial advice — educational framing only.
- Warm but not chatty. Under 250 words. No exclamation points.

Fill in your details and the prompt updates live — then copy.

What you get back (excerpt)

Great question — points confuse almost everyone at first. A discount point is prepaid interest: you pay more up front in exchange for a lower rate for the life of the loan. On your $320,000 loan, one point costs $3,200 and would lower your payment by about $54 a month. The way to judge it is break-even: $3,200 divided by $54 is roughly 59 months. If you expect to keep this loan longer than about five years, the point pays for itself; if you're likely to move or refinance sooner, you'd leave savings on the table. Happy to walk through it on a quick call.

The full workflow

  1. Pull the verified figures from the actual Loan Estimate or pricing engine.
  2. Run the prompt with the question and only those figures.
  3. Check the math and resolve any [CHECK] flags against current guidelines.
  4. Adjust to your voice and send from your compliant email account.

Watch out for

Program numbers change every year — FHA and conforming loan limits, MI factors, seasoning rules. AI training data is months to years stale; verify anything numeric against current guidelines before it goes to a borrower.

Never paste the borrower's Loan Estimate, pay stubs, or bank statements into a consumer AI tool — that's nonpublic personal information under GLBA. Retype the two or three verified figures you actually need.

Keep replies educational. The moment a draft says 'you should take the ARM,' it has crossed from explanation into advice that belongs to you as the licensed MLO, not to a chatbot.

Where this comes from

Every use case on this site is grounded in real reports from working mortgage brokers — not invented by us.

More AI use cases for mortgage brokers

← All 6 use cases: How Mortgage Brokers Use AI