Skip to main content

How Can AI Help Me Follow Up With My Real Estate Database?

Written by Coach Emily on . Posted in Uncategorized.

By Emily Terrell — Top Coach and Speaker at Tom Ferry International. Active San Antonio agent closing 70+ transactions a year.

AI helps you follow up with your real estate database by deciding who to call, researching why, and drafting a specific opener — so you make fewer, better calls instead of mass “just checking in” blasts. The human still makes the call. This guide covers the four-step workflow, the tools, and the compliance lines that keep it safe.

Key Takeaways

  • AI’s highest-value job in real estate database follow-up is preparation — who to call and why — not sending messages for you.
  • A real, recent, public detail beats “just checking in,” and AI can find one in seconds.
  • Cap each batch at 40 contacts so the system gets finished instead of abandoned.
  • Anything sensitive goes in a do-not-mention note, never in the opener.
  • Manual calls with AI-prepared notes keep you on the simplest side of TCPA, privacy, and fair housing rules.

What is real estate database follow-up with AI?

AI database follow-up means using an AI assistant alongside your CRM to:

  • decide which past clients and sphere contacts to reach out to,
  • find a real reason to contact each one, and
  • draft what you’ll say.

You still make the call. It isn’t a drip campaign. A drip sends the same message to everyone on a schedule. AI follow-up prepares a different conversation for every person on your list.

Why does database follow-up matter for real estate agents?

Here’s the thing nobody wants to tell you: the past client you helped years ago is probably only now nearing their next move. According to NAR’s 2025 Profile of Home Buyers and Sellers (November 2025), the typical seller lived in their home 11 years before selling, an all-time high (NAR). Your follow-up has to last a decade.

Good will isn’t the problem. According to the same NAR report, 91% of buyers would use their agent again or recommend them to others. People like their agent. They just don’t remember to call her when the moment comes, unless she’s stayed in touch.

AI isn’t closing that gap on its own. According to NAR’s 2025 Technology Survey (September 2025), only 17% of Realtors said AI had a significantly positive impact on their business, and 46% saw no noticeable impact (NAR). From what I see coaching agents every week, most of them point AI at content. Very few point it at the database, which is where the next ten years of their business sits.

“Your database doesn’t need more automation. It needs a reason for you to call. AI’s job is to find that reason before you pick up the phone — forty names, forty real openers, and zero ‘just checking in.’”
— Emily Terrell, Tom Ferry Coach

The four-step AI database follow-up workflow

This is the system. Each step has one job. AI handles the part of that job that used to eat your morning.

Who should you call this week?

Start with a list your CRM builds for you, then let AI rank it. In Follow Up Boss, Smart Lists work like saved searches. They filter your database by stage, last communication, and more, and they update automatically as you log contact (Follow Up Boss). Build one list of past clients and sphere you haven’t spoken to in 90+ days.

Then tier it:

  • A: people you talk to regularly and know the most about.
  • B: people you know something about.
  • C: names you don’t really know.

Give AI the A and B tiers with your CRM notes. Ask it to flag anyone with a recent signal: a closing anniversary, a renovation they mentioned, a timeline they told you about last time.

Cap the batch at 40. Forty prepared calls beat 300 you meant to make.

What should you know before you dial?

Know one real, recent thing about the person, and know what not to bring up. This is where AI saves the most time. Give it the contact’s CRM history and have it check their public social posts from the last 60 days. Start with Instagram, and use Facebook when Instagram is thin.

Two guardrails make this work.

Identity. Names aren’t unique. Require two independent signals before AI treats an account as the right person: city, employer, spouse, or how you know them. With only one signal, the contact gets marked UNCONFIRMED and you decide.

The hook test. Ask: would they be glad I noticed, or unsettled that I was looking? Life events and milestones pass. Anything sensitive goes into a separate KNOW-DON’T-SAY section, never the opener. That means health, divorce, job loss, death, finances, or politics.

Here’s the prompt structure I teach:

You’re helping me prepare database calls. For each contact below, use only the CRM notes I’ve pasted and their public posts from the last 60 days.

1. Confirm identity with two independent signals (city, employer, spouse, how I know them). If you only have one, mark them UNCONFIRMED.

2. Write one spoken opener, under 25 words, that references something real and does not mention real estate.

3. List anything sensitive (health, divorce, job loss, death, finances, politics) under KNOW-DON’T-SAY.

4. Record only what you actually saw. No guesses.

Plain text, under 200 words per person.

What do you actually say?

Say something that proves you remembered them, and don’t mention real estate. The opener has one job: start a real conversation.

  • “Saw the graduation photos — congratulations, that’s huge” works.
  • “Just checking in to see if you’re thinking about the market” doesn’t, because it asks them to do the work.

Have AI draft two openers per person, under 25 words, written to be said out loud. Pick one, adjust it so it sounds like you, and dial. If the conversation turns to real estate, it’ll be because they took it there.

Written follow-up after the call has its own rules. I cover those in my guide to AI for real estate client communication.

What happens after the call?

Log it the same day, while it’s fresh. Record a 30-second voice memo after each call and have AI turn it into a plain-text CRM note under 200 words. It should cover what you learned, any timeline they mentioned, and the next touch date.

Only record what they actually said. Speculation doesn’t belong in your CRM. A note like “probably thinking about downsizing” becomes a bad opener six months from now, and a worse look if the record ever gets read.

The note feeds next month’s Smart List, and the system runs again.

Where are the compliance lines?

AI-prepared calls keep you on the simplest side of the rules, but three lines still apply.

Client data. Call prep is different from listing marketing. The name is the point, so you can’t redact it away. That makes your account tier the control. Use an AI account that excludes your inputs from model training, and send only the minimum: name, relationship, recent context. Never send financial details or anything a client told you in confidence. I break down the account tiers in Is It Safe to Put Client Data in ChatGPT?

Automation. You dialing a contact with AI-prepared notes is manual outreach. AI sending texts for you, or an AI voice calling for you, is not. The FCC confirmed in 2024 that AI-generated voices fall under the TCPA’s artificial-voice restrictions and require the called party’s prior express consent (FCC). I walk through the consent tiers in AI Texting Rules for Real Estate Agents. Do-not-call obligations and your brokerage’s policies still apply to manual calls.

Fair housing. Prioritize your list by behavior and timeline: last contact, stated plans, engagement. Never prioritize protected characteristics or proxies for them. A life event can open a conversation. It should never decide who gets called.

This is general information, not legal advice. Confirm your setup with your broker and, where appropriate, an attorney licensed in your state.

How I use this in my own business

I close 70+ transactions a year in roughly five hours a week of active management. That math only works because my database does a lot of the lifting, and this is the system behind it.

My contacts are tiered A, B, and C:

  • A: people I communicate with regularly and know the most about.
  • B: people I know something about.
  • C: people I don’t really know yet.

My database calls start with research, not a script.

Before a call block, I have Claude work through a batch in my own browser. It’s never more than 40 people, because forty good notes beat 300 abandoned halfway. For each name, it reads my CRM notes, then checks public Instagram posts from the last 60 days, and Facebook when Instagram is thin. It needs two independent signals before it treats an account as the right person. If it only finds one, it marks the contact UNCONFIRMED and leaves the decision to me.

What comes back is a plain-text note, under 200 words per person, that pastes straight into my CRM:

  • a spoken opener at the top that never pivots to real estate,
  • a KNOW-DON’T-SAY section at the bottom for anything sensitive.

Nothing in it is a guess. The only things allowed in my CRM are things somebody actually saw.

Then I make the calls. Feet on the desk, coffee in hand, talking to people who are about to hear from someone who actually noticed them.

What should brokerage and team leaders take from this?

Database follow-up is the easiest AI workflow to standardize across a team. The tiers, the 40-contact cap, the prompt, and the KNOW-DON’T-SAY rule fit on one page. That means every agent can run the same system instead of improvising one.

Set the account-tier policy before anyone pastes a contact list anywhere. Make “notes only record what was actually said” a team rule. Both are easier to install in a training room than in a memo.

Common mistakes

  • Automating the relationship instead of preparing for it. An AI-written “just checking in” text sent to 500 people is still a blast. Your sphere can tell.
  • Building a list you’ll never finish. Three hundred names on Monday becomes zero calls by Thursday. Cap the batch.
  • Letting AI guess. If it can’t confirm identity with two signals, or it’s inferring something it didn’t see, that detail stays out of the note and out of the call.
  • Opening with real estate. “Thinking about selling?” ends the conversation before it starts.
  • Pasting a full CRM export into a personal AI account. That’s a disclosure you can’t take back.

Frequently Asked Questions

Can AI follow up with my real estate database for me?

AI can do most of the work around follow-up, but it shouldn’t make the call. Let it rank who to contact, summarize each person’s history, research public posts, and draft an opener. You make the call and log what you learned. Fully automated outreach to a sphere reads as automated, and automated texts and AI voice calls bring TCPA consent requirements with them.

What’s the best AI tool for real estate database follow-up?

The best setup is two tools working together: your CRM and a general AI assistant. A CRM like Follow Up Boss tells you who’s due through Smart Lists filtered by stage and last contact. An assistant like Claude or ChatGPT, on an account that excludes your data from training, turns each name into a research brief and an opener. The CRM handles timing. The assistant handles context.

How many database contacts should I call at a time?

Start with a batch you’ll actually finish. I cap mine at 40 people, because forty good call notes beat 300 abandoned halfway. Work your A contacts first — the people you talk to regularly and know the most about — then move to B. Consistency matters more than volume: a list you finish every week compounds, and a list you abandon in March doesn’t.

Is it legal to use AI to text my sphere?

Generally yes, when you’re the sender. If AI drafts a text and you read and send it yourself, that’s manual outreach. If an automated system sends on your behalf, TCPA consent rules apply, and marketing messages require prior express written consent. A past-client relationship isn’t consent to automated contact. This is general information, not legal advice; confirm your setup with your broker.

Is it creepy to research past clients on social media before calling?

Not if you use the right filter. Stick to public posts from the last 60 days, confirm you’ve found the right person with two independent signals, and run every detail through one test: would they be glad I noticed, or unsettled that I was looking? Anything sensitive — health, divorce, job loss, a death — goes in a do-not-mention note, never into the opener.

What should I say instead of ‘just checking in’?

Say something that proves you remembered them as a person. Reference a real, recent, public moment — a kid’s graduation, a new puppy, a finished kitchen remodel — and don’t pivot to real estate in the opener. ‘Saw the backyard finally got done — how’s the fire pit holding up?’ starts a conversation. ‘Just checking in’ asks them to do the work of making one.

Can I put my CRM export into ChatGPT?

Not into a personal account. On ChatGPT Free, Plus, and Pro, conversations can be used for model training unless you switch that off, while Business, Enterprise, and API accounts exclude your data by default. Even on the right tier, send only what the task needs: a name, the relationship, and recent context — never financial details, ID numbers, or anything a client told you in confidence.

Bring this to your team or event

Emily Terrell speaks at brokerage events, real estate conferences, and team trainings on AI, systems, and social media — the exact playbook in this post, delivered live to your audience. As a Top Coach and Speaker at Tom Ferry International and an active agent closing 70+ transactions a year, Emily speaks from the stage about what’s working right now, not theory. Recent stages include NAHREP and eXp Con.

Book Emily to speak at your next event:
Email: eterrell@yourcoach.com
Phone: (210) 400-9191
Web: coachemilyterrell.com

For real estate agents who want to implement this: Get the weekly real estate prompt library at weeklyrealestateprompts.com or follow @coachemilyterrell on Instagram for daily systems and AI breakdowns.