Welcome to AutomateRE, your weekly playbook for using AI in real estate.
CNBC just ran a story that every agent should read twice. Clients are showing up having priced their own home with ChatGPT. Ryan Serhant nearly lost a deal to AI-generated comps before he talked the client back to reality. His line stuck with me: "AI can model a market. It can't model a deal." It doesn't know intentions. It doesn't know what buyers are circling. It wasn't standing in the house.
That last part is the whole game. The agents getting beaten by AI right now are the ones who let the model guess: generic comps, generic listing copy, the same three sentences every other listing on the MLS already used. The agents winning with it feed it what they actually saw. So this week we're building the content version of that edge. You walk the house, talk into your phone for 90 seconds, and AI turns that one voice memo into your full listing description, social caption, email announcement, and MLS remarks. Not invented. Grounded in your walkthrough.
This is issue #20. Let's get into it.
Quick Win: One voice memo from the walkthrough becomes your listing description, social caption, email announcement, and MLS remarks
Record a 90-second walkthrough on your phone, paste the transcript, and get all four pieces of listing content in one pass, grounded in what you actually saw instead of what AI guessed.
Here's the prompt:
You are a listing agent's content writer. I just walked a property and
recorded a voice memo describing it. Below is the raw transcript. Turn it
into four finished pieces of listing content. Use ONLY the facts in my
walkthrough plus the listing details I give you. Do not invent features,
finishes, room counts, or amenities. If something I'd normally include is
missing, leave a [CONFIRM: ...] note instead of guessing.
LISTING DETAILS:
- Address / area: [STREET OR NEIGHBORHOOD, CITY]
- Price: [$X]
- Beds / baths / sqft: [N / N / N]
- Property type: [single family / condo / townhouse / multi]
- Standout fact buyers will care about most: [e.g. "new roof 2025" /
"walk to the train" / "primary on main"]
- Target buyer: [e.g. "first move-up family" / "downsizer" / "investor"]
MY WALKTHROUGH VOICE MEMO (raw transcript):
"""
[PASTE TRANSCRIPT HERE - just talk: the light in the kitchen, the
backyard, the street, the upgrades, the feeling when you walk in,
the one thing you'd tell a buyer in person]
"""
WRITE FOUR THINGS:
1. MLS DESCRIPTION (180-220 words)
- Lead with the single most compelling thing about this home
- Specific, sensory, true to the walkthrough; no "must see" filler
- Plain, factual fair-housing-safe language (describe the home, never
the buyer or who "belongs" in the neighborhood)
- End with one concrete line on location/lifestyle
2. MLS AGENT REMARKS (2-3 sentences, agent-to-agent)
- Showing logistics, what to flag to a buyer's agent, what makes it
show well; blunt and useful, not marketing fluff
3. SOCIAL CAPTION (Instagram/Facebook, 60-90 words)
- Hook in the first line, 3-5 relevant hashtags, one clear CTA
- A different angle than the MLS copy, not a copy-paste
4. EMAIL ANNOUNCEMENT (just-listed, to my database)
- Subject line + 120-150 word body
- Personal, like I'm telling my sphere about a home I actually like
- CTA to book a private showing before the open house
RULES:
- One voice, mine: [warm and direct / polished and upscale / fast and
punchy - pick one]
- No cliches: skip "nestled," "boasts," "must see," "won't last,"
"dream home," "in today's market"
- Flag anything you're unsure about with [CONFIRM: ...] rather than
filling the gap with a guessResult: A 90-second voice memo and five listing fields become four ready-to-publish pieces, MLS description, agent remarks, social caption, and a just-listed email, each built from what you saw in the house, with [CONFIRM] flags on anything the AI couldn't verify.
Industry Intel: The agents winning with AI are the ones who were actually in the room
Two recent stories point at the same line. AI is now everywhere in real estate, and that is exactly why generic AI output has stopped working. "Slop" was Merriam-Webster's word of the year for 2025, and buyers feel it: only 26% of consumers now prefer AI-generated creator content, down from 60% in 2023. The edge has flipped from "use AI" to "feed it what you actually know."
Ryan Serhant almost lost a deal to AI comps, and named the real moat
CNBC reported on June 16 that clients are increasingly pricing their own homes with ChatGPT and Claude before they ever call an agent. Coldwell Banker Realty CEO Kamini Lane said both buyers and sellers now show up with AI valuations, and that those general-purpose models miss the nuance of a home, a neighborhood, and a client. Ryan Serhant told the story of nearly losing a deal to AI-generated comps before walking the client back. His line: "AI can model a market. It can't model a deal." It doesn't know intentions, doesn't know what buyers are circling, doesn't know the off-market comp down the street.
The same gap shows up in your listing content. A listing description written from the MLS data sheet alone is the content version of an AI comp: confident, generic, and missing the one detail that actually sells the house. The model was never in the room. You were.
Your move: Be the input AI can't get on its own. Before you write a word, record the 90-second walkthrough from this week's Quick Win, the light in the kitchen at 4pm, the re-done roof, the quiet dead-end street. That memo is the nuance Serhant is talking about. Feed it in, and your copy reads like an agent who stood in the house, not like the other 200 ChatGPT listings on the MLS.
Listings are starting to talk back, and the content bar just went up
The National Association of Realtors profiled a shift on June 15: listings are moving from static digital brochures to interactive experiences. Startup Infinityy lets buyers ask questions, pull neighborhood insights, and connect with the agent by text, voice, or live video from inside the listing itself. CEO Lisa Nickerson framed the point well, that an agent joining a buyer "inside" the listing the moment interest happens beats any delayed follow-up.
Read that as a warning about content depth. When buyers can interrogate a listing in plain language, a flat three-line blurb has nothing to say back. The listings that hold attention are the ones rich with specific, true detail, and that detail has to come from somewhere.
Your move: Stop writing one thin description and moving on. The voice-memo kit produces your MLS description, agent remarks, social caption, and just-listed email from a single recording, so every channel carries specific, grounded content instead of the same recycled sentence. You feed the interactive-listing era instead of getting caught short by it.
Automate This: Voice Memo Listing Kit
Every week, we help you automate one part of your job. This week: a workflow that turns the voice memo you record during a walkthrough into your full listing content kit, MLS description, agent remarks, social caption, and just-listed email, before you pull out of the driveway.
The problem: You take a listing, you walk the house, and you notice everything that matters: the light, the upgrades, the one feature that will move a buyer. Then you get back to the office, the details have gone fuzzy, and you write the description from the MLS data sheet because that is what is in front of you. So the copy comes out flat. Then you rewrite it for Instagram, rewrite it again for the just-listed email, and retype the agent remarks. Forty-five minutes per listing, four times if you carry a real inventory, and the best detail, the thing you noticed standing in the kitchen, never makes it in.
The solution: We built you an automation workflow:
You finish the walkthrough and record a 90-second voice memo on your phone. Just talk: what you saw, the upgrades, the street, the one thing you'd tell a buyer in person. Drop it in a watched Google Drive folder, or send it to a dedicated address.
The new audio file triggers the workflow and gets transcribed automatically (Whisper / OpenAI), so your spoken walkthrough becomes clean text.
The workflow pulls that listing's details from a simple intake row you filled when you took the listing: address, price, beds, baths, sqft, the standout fact, the target buyer, and your voice setting.
Transcript plus listing details go through the Quick Win prompt above in one call, with the no-inventing rule and the [CONFIRM] flags built in.
AI returns all four pieces: a 180 to 220 word MLS description, 2 to 3 sentences of agent remarks, a social caption, and a subject line plus just-listed email.
Everything lands in a Google Doc titled with the property address, [CONFIRM] flags highlighted at the top so you can fix anything the AI was unsure about in 30 seconds.
You get an SMS: "Listing kit ready for [ADDRESS]." One tap to open, review, and publish. Everything logs to a Listings sheet so you can see which descriptions got the most showing requests.
After ten listings you stop staring at a blank MLS field and start editing a grounded first draft that already sounds like you. The detail you noticed in the house is the detail buyers read.
Time to set up: 40 to 50 minutes
Time saved: 45 to 60 minutes per listing
Want it? Reply to this email with "VOICELIST" and we'll send you the workflow file + setup guide.
Next week: an AI that writes your weekly seller update, showings, feedback, online views, and where the price sits, so your sellers never have to text you "any news?"
That's it for issue #20. Every agent has the same AI now, so the model is no longer the advantage. What you saw in the house is.
A 90-second voice memo from the walkthrough is the cheapest moat you have left, and it turns into four pieces of content that sound like you actually stood there. Record the next one before you pull out of the driveway, and reply if you want the automation.
- NextAutomation Team
PS: Know another agent, investor, developer or PM who'd find this useful ? https://newsletter.nextautomation.us/subscribe - it helps us keep this free.