A lead fills out a form on Tuesday. The agent is at a closing. By Thursday the lead has talked to two other agents. Nobody was lazy. The calendar just ate the afternoon.
Real estate is a follow-up business, and follow-up is the first thing that breaks when an office gets busy. That is where AI earns its keep: reminders, first drafts and sorting, so the human work happens on time.
This post covers what to set up, what to keep human, and the fair housing rules you need to respect when a machine writes anything about a home. For the big picture on adding AI to any small business, see our guide to AI implementation for small business.
Plain terms first
- AI assistant: a tool like ChatGPT, Claude or Gemini. You give it an instruction (a "prompt"), it writes a draft.
- Automation: a rule that runs on its own. New lead comes in, send a text, add a task, remind the agent in two days.
- CRM: the customer list where your leads and notes live. Many offices already have one and use only part of it.
AI writes the draft. Automation makes sure the draft goes to the right person at the right time. You need both, and you need a human in the middle.
The rule that comes first: a person reviews what AI writes
Anything an AI tool produces that goes to a client or the public gets read by a licensed person first. Listing descriptions, ads, social posts, emails, texts. No exceptions.
Two reasons.
Fair housing. The Fair Housing Act makes it unlawful to publish any advertisement for the sale or rental of a dwelling that indicates a preference, limitation or discrimination based on race, color, religion, sex, handicap, familial status or national origin. That is in 42 U.S.C. 3604(c). HUD's regulation at 24 CFR 100.75 goes further. It says the rule covers written and oral notices, and that discriminatory ads include using words, phrases, photographs, illustrations or symbols that convey a dwelling is available or not available to a particular group. It also covers selecting media or locations for ads in a way that denies a segment of the market information about housing.
Read that again with an AI writer in mind. A tool trained on a mountain of internet text can produce phrases that sound natural and still steer. "Perfect for young professionals." "Great for a growing family." "Quiet, adults preferred." A person who knows the rules catches those. A tool does not know it just crossed a line.
Accuracy. AI will invent details. A fourth bedroom that is really a den. A school zone that moved. A flood zone it never looked up. A listing is a representation about a property. You verify every fact against the MLS entry, the seller's disclosures and your own eyes.
This is general information, not legal advice. Confirm your office's advertising practices with your broker of record and counsel.
Where AI helps a real estate office
1. Instant first response, a human next
When a lead comes in from your site, a portal or an open house sign-in, the first reply should go out in minutes, not hours. An automation can send a short, plain text or email right away: thanks, here is who will reach out, here is when. That reply should be written by you ahead of time, once, and reviewed. It is not a place for AI to improvise.
Then the agent gets a task: call within the hour during business hours. The automation did the part that humans forget. The human does the part that closes. Our speed to lead guide covers the research behind this for the trades, and the logic carries over. The reply buys you the conversation, and the agent has to earn it.
2. Follow-up cadences that keep going
Many buyers are not ready on the day they inquire. A "cadence" is just a planned series of touches over weeks or months. Something like:
- Same day: quick thanks and a question about what they want.
- Day 3: a short note with two or three listings that match what they said.
- Day 10: a market update for the neighborhood they asked about.
- Day 30: a check-in that asks if their timeline changed.
AI can draft each message in your voice from a few of your own examples. Automation schedules them and stops the sequence the moment the lead replies or books a showing. An agent still reads the messages before the template goes live, and personal replies come from a person.
3. Notes, summaries and tasks after a call
After a showing or a call, agents are supposed to log notes. Most do not, because they are in the car. A recording or a voice note that an assistant turns into a three-line summary and a next step is a real time saver.
Two cautions. First, tell the client if you record a call, and check Florida's recording rules before you start (we cover this in our recording disclosure guide). Second, keep client details out of consumer tools that are not set up for business data. See is your business data safe in ChatGPT for a one-page policy.
4. Listing description drafts
This one saves real time and carries the most risk, so it gets a process:
- Give the assistant only facts you verified: bedrooms, baths, square footage, features, upgrades with years.
- Tell it to describe the property, not the buyer. No mention of who the home is "for."
- Ask for a short and a long version.
- A person reads every word against the fair housing rule above and the MLS data.
- Run the final through your broker's compliance step if you have one.
If a draft says anything about who would like the neighborhood, who lives nearby or what kind of people the area suits, cut it.
5. Transaction reminders
Contract dates, inspection windows, appraisal follow-ups, closing checklists. Missed dates cost real money. Reminders are a perfect job for automation, and AI can help draft the plain-language updates to clients. The deadlines themselves come from the contract and your transaction coordinator, never from the assistant's guess.
6. Sorting the old database
Most offices are sitting on past clients and cold leads. An assistant can help sort a list into "bought in the last five years," "inquired but never closed" and "sphere of influence" so each group gets the right message. That is where a lot of quiet revenue lives. Do it only with contacts who have a reason to hear from you, and honor opt-outs.
What to keep human
- Anything about price, value or negotiation. AI does not know your market this week.
- Anything that mentions who a neighborhood or home is for.
- Disclosures, legal language and contract terms.
- A reply to an upset client.
- Final approval of every public post and ad.
What to set up first
Pick one, not five. For most offices the biggest leak is lead response and follow-up after the first week. So:
- Time your own response. Submit your own web form at 8 p.m. and see what happens and when.
- Write the first reply and the 30-day cadence by hand, in your voice.
- Put it on automation, with a task for the agent.
- Add a review step before anything public goes out.
- Check in at 30 days: did more leads get a reply, did more showings get booked?
You can size the time you spend on these tasks now with the free AI time savings calculator.
Where Di-Hy fits
Di-Hy Digital is an AI implementation firm in Ybor City, Tampa. We start with a free audit, build in about 30 days, train your team, and offer monthly upkeep. For what that looks like in real estate, see our Tampa real estate page. If you want to talk first, book a strategy call.