AI prompts for follow-up scripts and texts

Written for real estate agents. Text-length follow-ups and objection scripts for the lead who went quiet — every message with a reason to exist.

• 01 Ten sections, one for each follow-up job a week actually hands you — from the first reply to the file you keep
• 02 30 fill-in prompts, brackets left open for your market, your notes, and how you sound in a text
• 03 30 follow-up steps — an audit that changes nothing first, then the call open and the short email built from the text you approved
• 04 30 named failure modes, each with the repair prompt you send in the same chat
• 05 A 118-page manual, a printable quick-reference page, and a plain-text file every prompt copies cleanly from

The first prompt in the kit works from two things you already have: the inquiry as it arrived and your own listing sheet. It tells the model to come back at text length — two sentences that fit under 320 characters — confirming one fact from the sheet, asking one question, and offering the single next step you named. Where the sheet is silent, the prompt tells the model to say so rather than fill the gap. Text length is the default output across the kit, not something you trim down to afterwards: twenty-three of the thirty fill-in prompts cap the draft in characters, twenty-one of them at 320.

The follow-up chain runs in the same chat. Step one is an audit that changes nothing — it tells the model to show quoted evidence for each claim in the draft, never to hand down a verdict on whether the message should go out. Only after you approve the text do seven of the ten sections turn it into the first line you would say on a call and an email under 120 words. The text stays the source; the longer forms are built from it, not the other way around.

Every section gives you three ways in: a prompt you fill in from your notes, a version that drafts in your own register — you paste a few messages you actually sent, stripped of anything identifying, and it matches how you write — and an interview prompt that asks you questions first, for the nights you cannot face a blank bracket. The ten jobs: the first reply in the first minutes, the note after a showing, the re-open for the lead who went quiet, the objections you hear weekly on the buyer side and on the seller side, past clients and the sphere that sends you referrals, the ladder of what goes out when and when to stop, the conversation you have been avoiding, the rewrite pass, and your follow-up file.

The thirty failure modes are most of what you are buying, and the first is the one every agent already knows: “just checking in.” Warm, polite, contentless — the eighth such note that person has received this year, with nothing in it about them and so nothing in it to answer. The repair makes the model open with one specific reason drawn from a note it can name — and when your notes hold no reason at all, say that plainly and draft nothing.

The rest are as recognizable once named. The re-open that makes the market the reason, or makes the silence itself the subject. The draft that answers a buyer objection with a forecast, turns the answer into a close, or goes after the other agent — and its seller-side cousins that argue with another agent’s number, sell the fee, or invent the season the market is supposedly in. The past-client note where the anniversary date is the entire message, or where the home has quietly become an asset with a value attached. The hard-news draft that buries the news under a friendly opening, diagnoses and prices a problem the model cannot see, or promises an outcome nobody controls.

The structural ones live in the ladder and the rewrite pass: a sequence with no end and no exit, five scheduled dates carrying one interchangeable message, a calendar that has replaced the reason. And the compression failure — a draft cut to your number that kept every topic and every ask in a telegram voice you would never use with a client you want to keep. Each of the thirty ships with its own repair, and the manual tells you which one you are looking at.

The spine of the kit is the no-message clause. Eight of the ten workhorse prompts end by refusing to invent a reason to write: when the notes hold nothing, the prompt tells the model to say so and produce no draft. The follow-up with no reason behind it is the one that costs you the relationship, and this kit would rather tell you that than manufacture one.

It drafts the message and never decides whether you may send it. Who may be texted or called, who sits on which list, and what a message must carry are your brokerage’s rules and your jurisdiction’s — the kit states no statute, no threshold, and no opt-out wording, and where a required line might exist it leaves a labeled blank and writes nothing into it. The privacy rule runs the same direction: the scripts work from what you remember and the notes you keep, never from a pasted contact record, and they describe the property rather than the people. No street address, no client name, no phone number or money figure, and never the reason behind a sale enters a chat window; every paste bracket names what to strip before you send.

The free starter card gives you the four rules and ten prompts that cut across every job, at principle level — enough to get a usable first draft out of a model. It does not carry this kit’s audit-then-convert chains, its thirty named failure modes with their repairs, or the quick-reference page. The prompts are plain instructions with no tool-specific syntax, designed to work in ChatGPT, Claude and Gemini; the test runs behind this kit were made on Claude only, and the manual says that plainly rather than implying a cross-model check that has not happened.

Real Estate Agent Kit 03 · $7 · 118-page manual · on sale now