The single biggest predictor of whether an open home visitor becomes a client is not how interested they seemed at the property. It is how quickly and how specifically you reach them after they leave. Most agents already know this. The thing that defeats them is the volume problem. Eleven names on a sign-in sheet, eleven different financial situations, eleven different reasons they walked through, and only one Sunday evening to do something about it.
The NAR 2025 Technology Survey found that 68% of agents now use AI in some form, and that 66% adopt new tech specifically to save time. Follow-up after open homes is where that time saving converts directly into commission, because every hour of delay between the property visit and the first personalised contact is an hour your competitors are using.
The triage decision before you write a single message
Before you open any AI tool, sort the sign-in sheet into three buckets. The same nurture treatment for everyone is the failure mode that makes AI follow-ups feel canned.
The first bucket is hot. They asked a buying-stage question. “Could we get a builders report this week?” “Will the vendor look at offers before Saturday?” These get a phone call inside 90 minutes and a written follow-up the same evening.
The second bucket is warm. They engaged with the property but flagged a constraint. Finance not yet approved, a current home to sell, a partner who wasn’t there. These get a personalised written follow-up that night and a 14-day nurture sequence.
The third bucket is cold but courteous. They came, they signed in, they didn’t ask anything substantive. These get one short, useful message. Push too hard here and you turn a name on a list into someone who unsubscribes.
This sort takes about three minutes if you scribble a tag next to each name on the sheet while it is still fresh. The AI cannot guess these tags, so do not skip the step.
Prompt one: turn rough notes into eleven personalised emails
Open ChatGPT, Claude, or Gemini on your phone and paste your tagged notes in. Quality of input is what separates a useful draft from a generic one. The prompt below assumes you have already written one or two scrappy lines per visitor. The brackets are the parts you change.
“I am a real estate agent following up on an open home today at [ADDRESS], a [BEDS]-bedroom home listed at [PRICE]. Below is my list of visitors with my notes on each one and a tag of HOT, WARM, or COLD. For each visitor, write a follow-up email under 100 words that references the specific thing I noted about them. Do not invent details I did not give you. For HOT leads, end with a direct ask for a private viewing or a chat tomorrow. For WARM leads, end with a soft check-in that acknowledges their constraint. For COLD leads, send one useful market observation about the area and offer to send new listings if they want. Use a warm, professional tone in New Zealand English. Sign off as [YOUR NAME], [BROKERAGE]. Notes follow:
[PASTE YOUR NUMBERED VISITOR NOTES]”
You will get eleven drafts back in roughly thirty seconds. Read every one before you send. The model can confidently invent details that were not in your notes (a kid’s name you never mentioned, a neighbourhood you never discussed) and that single hallucinated detail will cost you the lead the moment they spot it. If a draft contains anything you did not put in your input, delete that line.
Prompt two: build a 14-day nurture sequence for warm leads
Hot leads convert from a phone call. Cold leads convert from low-touch market emails. Warm leads are where the real money is and where most agents drop the ball, because a single email is not enough and ten emails is too many. The shape that works is three to four messages over fourteen days, each of them carrying one piece of useful information rather than one ask.
“I have three warm leads from today’s open home. For each one, draft a four-message follow-up sequence sent on day 1, day 4, day 9, and day 14. Each message must be under 90 words and offer something useful before any ask: a market data point, a neighbourhood insight, a similar listing, or progress on their stated constraint. Tone should sound like a knowledgeable friend, not a salesperson. Day 14 is the only message that asks for a meeting and only if they have not yet replied. The three leads are:
Lead 1: [NAME], [SITUATION], [WHAT THEY LIKED ABOUT THE PROPERTY], [STATED CONSTRAINT]
Lead 2: [NAME], [SITUATION], [WHAT THEY LIKED], [STATED CONSTRAINT]
Lead 3: [NAME], [SITUATION], [WHAT THEY LIKED], [STATED CONSTRAINT]”
Twelve drafted messages in under a minute. Schedule them in your CRM the same evening. The reason this works is that you wrote the strategy, not the AI. You decided who got nurtured. You decided what each lead’s stated constraint was. The model just typed faster than you can.
The counterintuitive part and the failure modes
Agents new to AI write longer follow-ups when they think they need to seem warmer. The opposite is true. A 70-word email that names the specific kitchen the visitor commented on reads as more personal than a 180-word email that name-drops three things they barely mentioned. A 30-word text on Sunday afternoon (“Hi Sarah, thanks for coming through 15 Beach Rd today. I noticed your reaction to the kitchen. The vendor is open to questions about the renovation history if you want me to ask. Cheers, James.”) often outperforms an essay-length email sent the same evening.
The first trap is letting the model fill silence with adjectives. Stunning. Beautiful. Exceptional. Vendors and buyers can spot that vocabulary at fifty paces. Add a line to either prompt: “Do not use the words stunning, beautiful, exceptional, perfect, dream, or opportunity.” That single constraint forces concrete nouns instead of marketing-speak.
The second trap is sending without reading. AI is fast enough that you can send eleven follow-ups in the time it used to take to write one. That speed is the gift. The risk is that you stop reading carefully, the model invents a detail, and you send confidently wrong information to a serious buyer. Two minutes per draft to read and check is non-negotiable.
What this changes about your Sunday evening
Before AI, an agent who wanted to do this properly was looking at three to four hours of writing on Sunday night. Most chose to send three or four good follow-ups and let the rest go cold. Now the entire list, hot through cold, can be drafted, checked, and scheduled in twenty to thirty minutes. The competitive edge is not that you have a tool the other agents do not have, since most of them now do. It is the discipline of triaging first, prompting precisely, and reading every draft before it goes out. That part still has to come from you.