AI for Loan Officers: 8 Claude Workflows That Save Hours

Most of the hours loan officers lose aren't on hard decisions. They're on rewriting the same explanations, follow-ups and recaps over and over. That's exactly the work Claude is good at.

The catch is that you work in a regulated business, with advertising rules, privacy rules, referral rules and a compliance department that reviews what you send. So this post starts with the guardrails, then gives you eight workflows with prompts you can paste into Claude today.

Nothing here is legal or compliance advice. Your company's policies win every time.

How do loan officers use AI without getting into compliance trouble?

Loan officers stay out of trouble with AI by following four rules: use only approved tools, keep borrower PII out, avoid rate quotes and trigger terms, and send anything public through compliance first. Here's the version I give in class:

• Approved tools only. Ask compliance which AI tools are approved, and for what.
• No borrower PII. Never paste Social Security numbers, account numbers, full credit reports, bank statements, or anything your company forbids into a tool it hasn't approved. Strip names and numbers and use placeholders like [BORROWER] and [LOAN #].
• No rates, no payment math. Mortgage advertising falls under Regulation Z. If an ad states a payment amount, the number of payments or repayment period, a down payment amount or percentage, or a finance charge, it can trigger additional required disclosures (such as the APR). Your compliance team knows the specifics. The easy fix: tell Claude to leave all of that out.
• RESPA Section 8 for agent co-marketing. RESPA generally prohibits giving or accepting a fee, kickback or "thing of value" for referring settlement service business, and "thing of value" can be read broadly. That's why co-marketing with agents needs a compliance check on who pays for what.

The CFPB publishes the regulations at consumerfinance.gov, but your compliance department is the final word. And Claude can make mistakes, so read every draft before it goes anywhere.

1. How do I use AI to write pre-approval follow-up emails?

Give Claude the borrower's situation in general terms and ask for a short sequence of follow-ups in your voice, with no rates or payment figures. Pre-approved buyers who go quiet are the easiest win in your pipeline.

When I teach this, the first question is whether the emails will sound like a robot. They will unless you paste in two or three emails you've actually sent and say "match this tone."

You are helping me, a mortgage loan officer, write a follow-up sequence for a borrower who was pre-approved but has gone quiet.

Situation: [e.g. pre-approved 3 weeks ago, touring with an agent, quiet for 10 days]
My tone: [paste 2-3 emails I've written]

Write 4 short emails to send over about 3 weeks:
1. A friendly check-in asking how the search is going
2. A helpful note about something they can do now to stay ready (keep documents current, avoid new credit, call me before big purchases)
3. A note offering to update their pre-approval if anything changed
4. A low-pressure "still here when you're ready" email

Rules:
- Under 120 words each, with a subject line
- No interest rates, APRs, payment amounts, down payment amounts, loan terms or other loan numbers
- Use [BORROWER FIRST NAME] and [AGENT NAME] as placeholders
- End each with one simple question

For a wider look at follow-up systems, see AI lead follow-up for real estate. The structure works the same for loan officers.

2. How can AI explain the loan process to first-time buyers?

Ask Claude to write a plain-English explanation at a specific reading level, for a specific moment in the process, with no rates or numbers. First-time buyers ask the same questions about pre-approval, closing costs and underwriting, and a clear written answer saves you a phone call.

Keep it generic. You're explaining how the process works, not what their loan will cost.

I'm a mortgage loan officer. Write a plain-English explanation for a first-time homebuyer about [TOPIC: what pre-approval means / what closing costs are and why they exist / the steps from application to closing / what underwriting is].

Audience: someone who has never bought a home. About 8th-grade reading level, short sentences, jargon defined in one line.

Format:
- A 2-sentence summary at the top
- 4-6 short sections with simple headers
- A "What you can do right now" list at the end

Rules:
- General education only. No rates, APRs, payment examples, dollar amounts or percentages.
- Don't promise approval, timelines or outcomes.
- End with: "Every loan is different. Call me and I'll walk you through your situation."
- Flag any sentence my compliance team might want to review.

Asking Claude to flag sentences for compliance won't replace the review, but it makes the review faster.

3. Can AI write a weekly market or rate-environment post?

Yes, if you frame it as education about what's happening and what it means for a buyer's planning, and skip the rate numbers and jargon. A weekly post keeps you visible with agents and past clients, and Claude can draft it in a few minutes from notes you give it.

Give Claude your own notes, not rates to quote, and run the draft through compliance.

I'm a mortgage loan officer writing a short weekly social post and email blurb for buyers and real estate agents.

My notes from this week: [paste 3-5 bullets in your own words, e.g. "rates moved around a lot," "more buyers asking about first-time buyer programs," "reminder: get pre-approved before touring"]

Write:
1. A social post under 150 words
2. An email version under 200 words

Rules:
- No interest rates, APRs, payment amounts, down payment amounts, loan terms or percentages
- No rate predictions and no jargon ("basis points," "MBS," "spreads")
- Focus on what this means for someone planning to buy, plus one practical step
- Note at the bottom which lines might need compliance review

4. How do loan officers co-market with real estate agents using AI?

Claude can draft co-branded content fast, like a joint first-time buyer guide or an open house flyer section, but the arrangement behind it needs to clear RESPA first. The writing is the easy part. Who pays for the flyer, the ad or the event is the part your compliance department needs to approve.

Compliance will look at the whole arrangement, not just whether money changed hands. Get the arrangement approved, then use Claude for the words.

I'm a mortgage loan officer creating co-marketing content with a real estate agent. Our compliance department has approved this arrangement: [describe, e.g. "shared first-time buyer seminar, costs split per our approved agreement"].

Write [a one-page buyer handout / an event description / a social post announcing the event] about [TOPIC].

Agent's role: [what they'll cover, e.g. finding a home, making an offer]
My role: [what I'll cover, e.g. getting pre-approved, understanding the loan process]

Rules:
- Educational, not salesy
- No rates, payment amounts, down payment amounts, loan terms or percentages
- Describe homes and the process, never who should live somewhere
- Leave placeholders: [MY DISCLOSURES] and [AGENT BROKERAGE DISCLOSURE]

If you want to share prompts with your agent partners, point them to Claude prompts for real estate agents.

5. How can AI help with post-closing check-ins and home anniversary notes?

Ask Claude for a short, personal note tied to a milestone (30 days after closing, the first home anniversary) built around one detail you remember. A note that sounds like you beats a generic "Happy Homeaversary" card every time.

I'm a mortgage loan officer. Write [a 30-day post-closing check-in / a one-year home anniversary note] for a past client.

What I remember about them: [one detail, e.g. "they were excited about the backyard," "first-time buyers," "moved for a new job"]
My tone: warm, brief, not salesy.

Write 3 versions: a text (under 40 words), an email (under 100 words), and a handwritten card (under 60 words). Use [CLIENT FIRST NAME]. No rates, refinance offers, payment amounts or loan terms. One light line saying I'm happy to help them or anyone they know.

If a refinance conversation makes sense, that's a call, not a template.

6. How do I turn an underwriting condition list into a borrower to-do list?

Strip the personal information out of the condition list, paste it into Claude, and ask for a plain-English checklist grouped by what the borrower has to do. A clean to-do list gets documents back faster than a raw condition list.

This is where the PII rule matters most. Remove names, addresses, loan numbers, account numbers and dollar figures first. If your company hasn't approved the tool for loan file content, skip this one.

Below is a list of underwriting conditions with all personal information removed. Turn it into a simple to-do list for the borrower.

[PASTE CONDITIONS WITH NAMES, NUMBERS AND ACCOUNT DETAILS REMOVED]

Format:
- Group under: "Documents to send," "Things to explain in writing," "Things I'm handling"
- One plain-English line per item; note which usually take longest
- End with: "Send these to me at [MY SECURE UPLOAD METHOD], not by regular email."

Rules:
- Don't add conditions or guess at amounts, dates or account details
- If a condition is unclear, list it under "Ask your loan officer" instead of interpreting it

Check the list against the original before sending. You're the one who knows what the underwriter meant.

7. Can AI turn call notes into CRM entries?

Yes. Paste your rough call notes (sensitive details removed) and ask Claude for a clean CRM note with next steps and dates. Loan officers I talk with love this one, because notes either don't get written or get written badly at 7 p.m.

Turn my rough notes into a CRM entry.

My notes: [paste, e.g. "talked w/ buyer, touring w/ agent sat, wants to stay under budget they mentioned, worried about closing costs, send closing cost explainer, check back thurs"]

Format:
- Contact type, a 2-3 sentence summary, and what they care about
- Next steps, each with an owner (me or them) and a date
- A short follow-up text I can send today

Rules:
- Replace any SSNs, account numbers, income figures or credit scores with [REMOVED]
- Don't invent details I didn't write down

If you record calls, confirm your company's policy on recording and on using transcripts with AI before you paste one in.

8. How do I write a first-time buyer FAQ page with AI?

Give Claude the real questions your borrowers ask and have it draft short, plain answers for a website or handout, then send the whole draft to compliance. Pull the questions from your own inbox. Those are what real buyers type into Google.

I'm a mortgage loan officer building a first-time buyer FAQ page for my website.

Questions my borrowers ask: [paste 8-12, e.g. "What's the difference between pre-qualified and pre-approved?"]

For each, write a 2-4 sentence plain-English answer and one "good to know" tip.

Rules:
- General education only. No rates, APRs, payment examples, dollar amounts, percentages or loan terms.
- Don't promise approval or timelines. Skip program eligibility details; say "ask me which programs fit your situation."
- Add [NMLS ID] and [REQUIRED COMPANY DISCLOSURES] placeholders at the bottom
- List any answers compliance should look at closely

How do I set this up in Claude so I'm not starting from scratch?

Set up Claude once with your rules and voice, and every chat starts already knowing them.

Create a Project with your disclosures and voice

A Project is a workspace with its own instructions and uploaded files that every chat in it uses. Make one called "Borrower Communication" and put this in the instructions:

I'm a mortgage loan officer. In every draft:
- No interest rates, APRs, payment amounts, down payment amounts, loan terms or finance charges
- Never ask for or repeat SSNs, account numbers, credit details or income figures
- Plain English, matching the tone of my uploaded sample emails
- Put [REQUIRED DISCLOSURES] at the end of marketing pieces and flag anything for compliance review

Upload your approved disclosure language and a few sample emails, only what your company allows.

Turn your follow-up style into a skill

A skill is a reusable playbook that teaches Claude one specific workflow. You can build one by describing your process in a chat ("here's how I follow up with pre-approved buyers"), and Claude builds and packages it. Turn it on in Settings > Capabilities > Skills, and Claude uses it when it's relevant.

Connect Gmail and Calendar, if your company allows it

Claude's Google Workspace connectors can read and search Gmail, draft replies, and check or create Calendar events. Gmail asks your approval before sending by default. On a Team or Enterprise account, an owner has to enable connectors first. Many lenders restrict email access, so ask before connecting anything.

Schedule a Monday pipeline briefing

Scheduled tasks (paid plans) let Claude run the same job on a cadence, and a task that reads files on your computer needs the desktop app. A good one: every Monday morning, read the pipeline report you exported from your loan system and list who needs a call this week and why. Claude works from the file you give it, and only use this if your company approves that data in the tool.

Every Monday at 7 a.m., read the newest pipeline report in [FOLDER]. List loans closing in the next 14 days with open items, and pre-approved borrowers with no activity in 2+ weeks, each with a suggested first-line text. One screen. No account numbers or personal financial details.

New to Claude? Start with how to use Claude for real estate agents; most of it applies to loan officers too. I also put together the free Claude 101 kit for agents and loan officers, with a class replay and a one-page cheat sheet.

What should loan officers do this week?

Pick one workflow and get it working before you add more. Here's the order I'd go in:

1. Ask compliance two questions: which AI tools are approved, and what data can go into them.
2. Build the Project with the instructions block above and your approved disclosures.
3. Run workflow 7 (CRM notes) after your next three calls. It touches the least sensitive data and saves time right away.
4. Draft one follow-up sequence for a pre-approved buyer who went quiet, and send it after you've read every word.
5. Write down your follow-up style and turn it into a skill once you've used the prompts a few times.

The loan officers who get the most out of this aren't the most technical. They set up their rules once, then use Claude daily for the boring writing.