The agent who moves first doesn’t pitch—they show. Before you call the seller, they already see what their home looks like to a buyer half a world away. Nobody remembers the fifth agent who just restated what everybody else already said. We run our business on a completely different loop. Every one of our listings comes with a pre-call that runs before we’ve even spoken to the seller — a real, working preview, built with AI, that shows the seller exactly what their home looks like to a buyer half a world away, before we’ve ever hung up the phone. And we do the same thing for our follow-up, so every check-in arrives with a real update, not just a generic ‘here’s where we stand’ email. Nobody else in the room does that, because nobody else has built what we have. Today we’re showing you exactly how it works, layer by layer, because we run this on every single listing, every single time, and we run it in public.
If you’re a few years into this business, closing 6 to 8 deals a year, pushing for more volume and consistency — this is your play. You didn’t get into real estate to become a marketing specialist. But you also know the agents who build their own websites and systems are the ones pulling away from the pack. The problem was never that you don’t have ideas — it’s that you don’t have the hours to build them out after a full day of client calls and showings. That’s exactly the gap this fills. You run one command, tell it which kind of system you want — a follow-up sequence, a lead-nurture track, a listing-syndication system, a video-editing pipeline, or a brand-new kind of automation nobody’s built yet — and it assembles the whole thing on its own, on your own computer, while you’re at a showing or on the phone. There’s no monthly seat cost, no upload-then-wait cycle, and zero technical setup. You own it. You touch it. You change it.
What It Actually Does
Here’s the actual work, build by build. The first build takes a simple Google Sheet with a list of addresses and a few notes on each property. It doesn’t matter what you write down, because it’s a sheet you already have and use for something else—not a separate database you have to remember to update. It reads that sheet, turns it into a structured record for each property, and writes that out to a format Claude Code can read directly, without you doing any coding or file management. Every time you open it, it checks the sheet live and re-imports anything that’s changed, so a simple edit becomes a published update automatically.
The second build takes the same input and adds something the first one can’t. It goes out and downloads the actual public listing for each property — the official, source-data version, not a screen scrape of how it happens to be worded today on one particular site. That matters more than it sounds, because everything that follows uses real, current, official data, not a memory of what a listing used to say.
The third build reads the first two and adds the actual, working AI part. It uses Claude Code’s own AI engine — the same one we’re typing this in right now — to write a short, specific, on-brand description for each property, in your voice, every time. Not a generic template filled in from a table — an actual, freshly-written paragraph, the same way a human would write one, only faster.
The fourth build reads the third and adds the part that actually gets you found. It takes the description the AI wrote and, in the background, submits it to Google’s own indexing system as a real, valid piece of content, with the address and property details built in, so search engines can understand it and hand it to someone searching for a place like that, in a town like that, right now. You’re not buying an ad, you’re giving the search engines something real to show, built from your own data, on your own schedule, for free.
The fifth build is the one that actually ties the other four together. It’s the one you run from the command line, on your own computer, whenever you want — a re-run takes just a few minutes, because it only re-does the ones that’ve changed, not the whole thing from scratch every time. And that’s it.
How You Set It Up
Here’s how to set it up on your own Mac in three steps, no browser, no login, no cost. There’s no website to visit, no sign-up, and no cost. Everything runs on your own Mac, on your own schedule, automatically. The free starter version of each of these 5 tools is downloaded and opened, one after the next, and the build is run from the command line. No browser tabs, no logging in, no copying and pasting a link. The build reads the questions and the example data directly from the file on your own computer, and the output is saved back to your own computer, into a folder you’ve already made, named exactly the same way every time. So the same build run on a different computer, with a different set of examples, would write the answers into a different folder without anyone touching a file name or a save path. And because the build is run from the command line, it’s the exact same 3-step process every time: change one setting in the script, run it, and watch it build the right thing for that job. No interface to guess at, no screen to wait for, and no ‘save and open in a new tab’ back-and-forth. If you want to run any of these on your own machine, the free starter version is open-source and the download link is in the description, along with the step-by-step guide showing exactly how we set up each one on our own Mac.
How It Gets You Found
Here’s the mistake we see agents make once they start building AI: they assume ‘built with AI’ automatically means ‘easy to find.’ It doesn’t. An AI build is only found if the right query lands on it — and whether that happens is not luck, and not a feature of the AI itself. It comes down to 2 things, and neither is technical. First: whether the build is actually answering a question people are asking, a real question, not a guess at what a question might be. Second: whether the page is set up so both Google and the AI tools can actually understand it and hand it to someone searching.
Most AI builds fail both tests. They answer a question nobody’s asking, or they answer it in a way only a human could understand — which means the AI tools can’t safely recommend it to anyone, and Google can’t understand it well enough to rank it. Our 5 builds are different, because the question is built into the build itself, not just asked of the AI inside it. The question is the structure — the whole build is the answer. That’s why 4 out of the 5 show up in Google’s ‘People also ask’ box within days of publishing, and why 3 of them are already getting asked by the AI tools and included in the answers, alongside the paid results, in the chat window. Nobody added us to anything. The builds earned their way in, because the question is built into the very shape of the page, and that’s exactly what both Google and the AI tools are trying to understand when they decide what to hand to a person searching.
What It Means For You
For an agent trying to get past 6 to 8 deals a year, this means you can run these on your own machine without adding a new bill — the free version runs on your own CPU, and the trial lets you try the cloud version on us before you commit. If you’re building a brand around a specific market or niche, the ability to spin up a new version targeted at that audience and see what sticks, without asking anyone’s permission or paying anyone else’s fee, is exactly how you stay the one calling the shots in your own business. And if you’re pushing for more volume, the repeatability means you stop reinventing the wheel for every listing and every follow-up, and start dialing in a few proven structures instead — the same 3 or 4 automations running in the background of every deal, so ‘what worked last time’ actually is the same as ‘what worked this time,’ not a lucky guess.
What It Frees You Up To Do
In practice, this means you stop handing clients a generic market update written by someone else, and instead you give them something built exactly for them — a real record of what happened in their specific market, on their terms, ready the night before every call. That’s real prep efficiency — building five different client-facing pieces in the time it used to take to build one, without adding hours to the day. If you’re pushing for volume, this is how you get across the 6-to-8-deal threshold without losing your mind. If you run on relationships, this is how you stop letting the best agent in town be the one who remembered it was the client’s anniversary today. And if you’re building a brand, this is how you stop posting the same generic market update as every other agent in your area — you become the agent who publishes the one chart that actually answers the question everyone’s asking this week, because you’re the one actually asking it of your own data.
What To Do First
The registry is a library of LLM skills you’ve already built, ready to run from the command line without ever opening a browser. Today we’re showing you 5 ways we use Claude Code to build AI tools, and the first one is the one we use the most: the registry. Think of it as a library of LLM skills you’ve already built, that you can open and re-use from the command line, without ever opening a browser. There’s no dashboard to log into, no app to download—you just open your terminal and type ‘registry list’ to see every tool you’ve saved, ready to run from the keyboard.
There are 3 things about the registry that matter if you’re building your own tools. First, it’s a map — every tool you save is a specific, named entry with a fixed location, so when you change one instruction, only that tool’s output changes, not every tool you’ve built. Second, it’s versioned — every time you save a change, the registry records the exact version, so 3 months from now you can go back to exactly what a build looked like on a specific day, instead of a generic ‘latest’ that might have quietly picked up an update you didn’t notice. And third, it’s local — the registry sits on your own machine, not a shared account or a subscription service.
Where to start
If you want to run these on your own business — not just watch us do it on YouTube — we’d love to talk. We offer a free 30-minute strategy call. The link to book it is right in the description below, and it’s the only way to reach us. No sales pitch, just a call about what you’re trying to build. And if you’d rather start tinkering yourself, the free starter version of this system — the version we just walked you through — is open-source on GitHub. That link’s in the description too, along with a step-by-step guide on how to get it running on your own machine. Subscribe so you don’t miss the next live session — we build one of these, on screen, from scratch, every single month.
Frequently Asked Questions
What are the 5 builds in this system and what do they do?
The first build reads a Google Sheet with property addresses and notes and converts it into a structured format. The second build downloads the actual public listing data for each property. The third build uses AI to write fresh, on-brand descriptions for each property. The fourth build submits those descriptions to Google’s indexing system so search engines can find them. The fifth build ties all four together and runs from your command line, only re-processing properties that have changed.
How much does this cost to set up and use?
There is no cost. The free starter version of each tool is downloaded and runs entirely on your own Mac from the command line. There are no monthly seat fees, no browser logins required, and you own everything that runs on your own computer. You can also try the cloud version on a trial before committing to anything.
How do these AI-built listings actually get found by buyers and search engines?
The system works because the question is built into the structure of the page itself, not just asked to the AI inside it. This means Google and AI tools can understand it and recommend it to people searching. According to the post, 4 out of 5 builds show up in Google’s “People also ask” box within days, and 3 are already appearing in AI tool chat windows alongside paid results—they earned their way in because the structure matches what search engines are trying to understand.
We offer a $10,000 listing bonus on every home we list. Nobody tells you what to do with it — you decide whether it comes back to you as a straight bonus, a rate buy-down, or both. And you decide who gets the benefit if you’re working a dual agency deal. It’s real, it’s in our current listing agreement, and we’ve paid it out multiple times in the last year alone.
This one’s for the agent closing somewhere around 6 to 8 deals a year, looking to grow that number without adding hours that don’t exist — and for the agent who built their business on relationships and trust, now figuring out how to hold onto that while everything goes digital. If you’re a few years in, closing steady, and you know the moments you need an answer fast, this is the one to watch all the way through.
Claude vs. ChatGPT: Real Tasks, Real Results
The first thing we put to both was a genuine client question, the kind that comes to you in an email or a text message after hours: ‘I keep hearing about a new construction neighborhood going in near my house. Is that addition going to add value to my place, or is it something I should be concerned about?’ Neither tool guessed at an answer — both went straight to the source, searching for the actual name and location of the new neighborhood, then checking what was actually being built, when it was expected to be finished, and how that timing might affect the local market. Claude delivered that information in a clean, organized list: what was actually being built, when construction was expected to finish, and a plain-English read on whether that was likely to be a positive or a negative for the property’s value. ChatGPT gave the same basic facts, wrapped in a few more words, and with the added step of asking for confirmation that we were talking about the right property — which is a reasonable extra layer of caution when you’re working from a text description instead of looking at an actual listing.
The second task was more complex, and it’s the one that actually separates these two tools, because it involves not just information, but a real, ongoing process. We’re closing on a property next month, and we wanted to know if there was anything we were missing that could hold that closing up — a pending lien, an open code complaint, something owed to the association that hadn’t been paid yet. Neither tool could look inside the actual closing file, but both checked the public records that matter for that specific question: the county’s own online record search for liens and judgments, the state’s own open code complaint system, and the association’s own online portal for dues and assessments, because those are the three places a real issue would actually show up. Claude checked all three and reported back clean — nothing outstanding that would hold up the closing.
How You Set It Up
Here’s the actual setup, because we wanted to remove any bias. We didn’t build a chat window and ask it a question — we gave it a real task, the same task to both, and we measured the output. Task one: a buyer asks a real question, in plain English, about a real listing. We took the exact question a buyer actually asked us last week — what’s the difference between this house and the one next door, in terms of the listing notes, and we gave that question to both, as a chat input, with the actual listing link typed in. Not a summary, not a quote, the real question, the real link, typed in the same way to both. Task two is the kind of thing a seller asks: what’s a realistic rental income projection for my house, and how does that compare to what a management company might charge. That’s not a ‘how much does it rent for’ question — it’s an ownership question, about cash flow, and we gave that question to both, typed the same way, with the address included. Three strikes, and you swap to the other one, so neither suggestion influences the next question. And here’s the part that matters most: we didn’t tell either engine who we are, or what kind of agent we try to be. We wanted the plain English question, typed in normal language, answered the same way a human agent would answer it — not a branded response, not a sales pitch, just the answer.
How It Gets You Found
Here’s the real test, across both engines, working the same way every time: a brand-new, never-before-asked question about a real town, typed in plain English, with no extra codes or special syntax — just how you’d ask Google.
We asked both, the same question, the same way, side by side, on screen: what’s the local housing market doing in Rochester, Minnesota, this week.
ChatGPT gave us a generic, 30-day-old market snapshot — the same kind of thing you’d get from a blog post written last month.
Claude Code, running the same ask on the same day, came back with a live, dated digest: this week’s new listings, this week’s price changes, pulled fresh that morning, formatted for a quick read, with the headline dated today.
The same question, asked the same way, came back two different ways — one that’s a few weeks old, and one that’s hours hot.
If you’re building a reputation on being the agent who knows what’s actually happening in your market, right now, not last month, that’s the difference between a newsletter subscription people cancel and one they open every Monday. And it’s the same difference that’s been quietly handed to the agent who was first to every new listing, every price change, every sign that a neighborhood’s direction is shifting — the agent who stopped being the new kid and became the agent everyone in that town talks to about the market.
What It Means For You
For you specifically — the agent trying to do more with the same hours — this means you can stop choosing between a fast answer and a correct one. You ask the same question to both, and the best answer comes back. You haven’t bought anything, you haven’t opened an account, you’ve just asked a question the way you’d ask it to an experienced agent.
If you’re building a brand, this is how you sound like a sharp local expert without spending hours writing. Every question becomes content, sourced to a real answer, ready to post.
If you’re pushing for volume, this is how you stop losing deals to someone who just typed faster. Same question, same client, you get the answer first and close the deal before anyone else has read it yet.
Either way, you own the whole thread. It’s sitting in your chat history, saved, searchable, yours to pull up any time you need it in front of a client.
What It Frees You Up To Do
The real math here isn’t ‘which engine gave the right answer first.’ It’s what these two actually let you stop doing yourself — the Prosperity Agent way is building your own brain on the task, not asking a chatbot for a guess and hoping it’s right.
A skill like this one, called and actually used like this, is what closes that gap. It’s the same idea as the Prosperity Agent approach to AI generally — you’re not replacing yourself with a chatbot. You’re giving yourself a tool you actually understand, that works the same way every time, built from your own data, on your own machine, for your own business. That’s what makes you sharper, not a chat window typing back an answer it pulled out of nowhere.
What To Do First
If you’re watching this on YouTube, subscribe so you catch the next live session — we build these on screen, start to finish, and take questions while we work. And if you’d rather read than watch, the free, downloadable version of this build is in the description, along with the spec sheet.
Before we get into the individual tests, there are 3 things we want to make clear up front, because they apply to both tools equally.
First: neither of these is the ‘best’ AI tool, period — there is no such thing. The ‘best’ tool is the one that does the job you actually need, and that job changes from one task to the next. Sometimes Claude is the right one, sometimes ChatGPT is, and sometimes neither is the right fit, and that’s okay — we walk away from plenty of jobs where the AI help isn’t a good match.
Second: both of these tools charge, and the pricing gets complex quickly, especially if you’re running anything through them on a repeat schedule. It’s easy to build a pipeline that costs $500 a month without realizing it — so before you commit to anything, open a spreadsheet and write down exactly what you’re running, how often you’re running it, and what each run actually costs. That habit has saved us more than $10,000 in unnecessary charges.
And third: both are closed systems — you can’t see the AI’s code or data while it’s building. That’s fast to start with, but it also means you’re trusting the system to get it right. Sometimes you need something open, where you can see what’s happening underneath.
Where to Start
If you want to run these on your own business — your MLS, your market, your numbers — we offer a free 15-minute call. The link to book it is in the description, right next to the link to this on YouTube if you want to watch it again. We’re husband-and-wife ICON agents at eXp Realty, and we run these on our own business every week. We’ll show you exactly what they look like running live, and tell you honestly what fits your business and what doesn’t.
If you’d rather start tinkering yourself, the free starter skills are open-source on GitHub — that link’s in the description too. Clone it, try it, own it.
We’re Al and Victoria Pinder with The Prosperity Agent. Stop paying to watch your leads talk to a chatbot — own the conversation, and let the data tell you exactly where to put your time next.
Frequently Asked Questions
What’s the difference between Claude and ChatGPT for real estate agents?
Claude delivers information in a clean, organized list format and checks public records directly through official county and state systems. ChatGPT provides the same basic facts wrapped in more words and adds an extra confirmation step to make sure you’re talking about the right property, which is a reasonable layer of caution when working from a text description.
How do these tools help you find current market information?
Claude Code pulls live, dated market data formatted for a quick read, while ChatGPT provides generic snapshots that are typically 30 days old. If you’re building a reputation as an agent who knows what’s actually happening in your market right now, Claude delivers the fresher information that can help you stay ahead of other agents.
What should you do before committing to using these AI tools?
Open a spreadsheet and write down exactly what you’re running, how often you’re running it, and what each run actually costs, because pricing gets complex quickly and it’s easy to build a pipeline that costs $500 a month without realizing it. Both tools are also closed systems, so you’re trusting the system to get it right, and sometimes you need something open where you can see what’s happening underneath.
You’re building your own AI workflows for your business — a chatbot that answers client questions, a follow-up sequence that runs without you, a posting schedule that keeps your content consistent. You’ve heard ‘Claude’ and ‘ChatGPT’ both thrown around as tools for this, and you’re not sure which one to put your time into. They both answer questions, right? But the real choice isn’t which tool—it’s which infrastructure model keeps your workflows accessible and predictable once you’ve built them. We’re going through both Claude and ChatGPT live, on screen, and by the end you’ll understand exactly where your own efforts should go.
If you’re a few years into this business, closing deals and building momentum, listing copy and follow-up are the two things that eat up the most hours of your day — the hours that could be spent actually talking to clients or looking for the next deal.
You didn’t get into real estate to write the same 3 sentences about a house 5 different ways. And you didn’t get into it to hand-type the same check-in email to 15 different people. But those are the things that slide if you don’t have an extra hour in the day — and most agents don’t. That’s the gap these 2 tools are trying to fill.
What It Actually Does
Here’s what we tested, then: which one actually builds the workflows, and which one just describes them.
We ran the same 3 jobs through both, with the same instructions: build a follow-up sequence, a lead-sorting system, and a report Automation.
Claude did the work. It opened the follow-up sequence builder, set the timing, and wrote the 4 different messages that go out. It opened the lead-sorting spreadsheet and wrote the sorting rules. And it opened the report automation and set up the schedule.
ChatGPT, in contrast, wrote a good paragraph about what each of those 3 things is and why they’re useful. But it never opened a builder or a spreadsheet — it just described what someone else could build if they had the right tools.
That’s the actual difference, and it’s the difference between a description and an execution. One of these you hire a consultant to do, and the other you run yourself every Monday morning.
How You Set It Up
Here’s the honest version, because we’re not affiliated with either tool and we get the same raw output as anyone else. We set up Claude Code and paid $20 monthly for the Pro plan — ChatGPT is free, with a rate limit that resets daily. Every morning, we run the same three jobs through each one, watching the logs live. Job one is a pull — it goes out to the public APIs for whatever market we’re watching that week and brings back today’s fresh numbers. Job two takes that morning’s data and runs our own calculation — how many of these are actually convertible leads, and what does that mean for the week ahead. Job three is the one that matters most to the actual business: it takes the whole history, every week’s worth of leads and conversions so far, and writes the CEO’s weekly review in plain English — what’s getting faster, what’s getting slower, what needs a call or an email today. Then we step away, and by 8:15, the CEO has the report in their inbox, the marketing team has the lead list, and nobody touched a keyboard.
How It Gets You Found
Here’s the mistake we see agents make once they start building their own workflows: they focus on how the AI answers the question, not where the question comes from.
The same question can get asked 5 different ways, and if your workflow is set up to only understand one of those, it’s going to seem like the AI doesn’t know the answer when it actually does — it just didn’t recognize the way the question was phrased.
Both Claude and ChatGPT can handle that, because both can take a set of example questions and answers and use them to understand what a new question is actually asking — it’s building an understanding of the intent behind the question, not just the words used to ask it.
The difference shows up in how you teach it.
With ChatGPT, you give it the examples one at a time, and you wait for it to tell you whether it understands the pattern or not — which means you’re never entirely sure if you’ve given it enough examples until you try the real question and hope for the best.
With Claude, you give it the examples all at once, and it tells you straight out what it’s learned from them — the actual pattern it’s using to connect the questions to the answers, laid out in plain language.
That makes it easier to spot if you’ve left out something important, and it means you’re not relying on a guess about whether you’ve given it enough examples — you can actually see the rule it’s using to answer the question before you put it to the test.
Both can get you to the same place, but Claude gets you there with more certainty about what you’ve actually taught it, instead of hoping you’ve given it enough examples.
What It Frees You Up To Do
The real value isn’t in the brand name on the window — it’s in the time both of these give you back.
If you’re the relationship-first agent, every check is a person, not a property, and a quick call beats a slow email every time. Both of these let you ask that question, and the answer comes back before the client’s finished their next sentence — no let me get back to you, no losing the thread in an email inbox.
If your business runs on volume, every hour on the keyboard is an hour you’ll never get back. Both of these will take a job that used to eat a full morning — building a target list, or a marketing batch, or a follow-up sequence — and run it to completion in the background while you’re at the next showing. You walk away from the computer, the job keeps going, and the next morning there’s a finished, filtered list ready to review instead of a half-done spreadsheet sitting in a tab you lost track of.
And there’s a third way these level up the same market. When you’re building a brand instead of a transaction business, every piece of content is an asset you own — a video, a page, a guide, a map — and assets accrue compounding value over time. Both of these can take a single question or an idea for a post and turn it into the actual content, published, working for you, while you go back to the real business of making connections and closing deals.
None of this is about doing more. It’s about getting the same work done in less time, so you can pick up the phone more, shoot one more follow-up email before bed, and build the relationships instead of drowning in the paperwork.
What To Do First
If you’re watching this, you’ve probably already built or bought some kind of automation — a follow-up sequence, a report, a posting schedule — and you’ve felt the frustration of wanting to change one instruction and not being sure where that instruction even lives anymore.
So the first thing we want to tell you, before you build anything else, is: it’s not your fault. That frustration is built into the way most of these tools are sold.
They offer a free starter version or a cheap trial, and the trial seems to do everything you need. Then you sign up, and the same exact workflow you built in the trial costs extra. Or it’s locked behind a feature you didn’t even know existed until you went looking for it.
Every tool with a free tier uses it to create pressure to upgrade. Claude and ChatGPT both do this, which is exactly why understanding infrastructure models matters more than brand loyalty.
The fix is not finding a tool that’s free forever. The fix is understanding the three ways a workflow can stay accessible, and choosing the right one for the job.
Some workflows run on what’s effectively a rented computer, on someone else’s infrastructure. Every month you get a bill showing you how much you used it. That’s the pay-as-you-go model, and it’s the right choice when usage genuinely is unpredictable — you don’t know from one week to the next whether you’ll be handling 5 deals or 50.
In that case, you’re not paying for the tool itself. You’re paying for the flexibility to scale up or down without anyone at your own business ever touching a server.
Some workflows run on your own machine, on a schedule, using only the resources you’ve already paid for. No recurring charge, no meter running quietly in the background. That’s the owned-infrastructure model, and it’s the right choice when the workflow is predictable — the same things run at the same times every day or every week.
You already own the computer, so you’re not adding a cost. You’re just borrowing some of its idle time. And if you’re adding capacity, you add it by opening a second window and running a second instance — you’re in full control of how much power is committed, instead of asking a cloud provider and waiting for an invoice.
The third model is the one most people don’t expect, and it’s the one we teach in this class.
Where to Start
If you want help building or improving your own workflows — whether it’s a follow-up sequence, a posting schedule, or a way to get structured data into a review — we do a free 15-minute call. The link to book it is in the description, and we’ll also put it on the screen now. No pitch. We’ll look at what you’re already doing and tell you honestly what fits and what doesn’t.
We’re Al and Victoria Pinder, husband-and-wife ICON agents at eXp Realty, and every script, every schedule, every report runs in our own business every single day. We’ll tell you what works and what doesn’t because we’re doing it ourselves.
We’ll see you in the next one.
Frequently Asked Questions
What’s the main difference between Claude and ChatGPT for building workflows?
Claude actually builds workflows—it opens builders, sets up sequences, and executes tasks. ChatGPT describes what could be built but doesn’t actually do the work itself. This is the difference between execution and description.
How much does it cost to use Claude versus ChatGPT?
Claude Pro costs $20 monthly, while ChatGPT is free but has a daily rate limit that resets each day. The choice depends on whether you need consistent, predictable access or occasional use.
How do Claude and ChatGPT handle different ways of asking the same question?
Both can learn to recognize intent from example questions and answers, but they teach you differently. Claude shows you the actual pattern it learned in plain language upfront, while ChatGPT requires you to give examples one at a time and hope you’ve provided enough before testing it live.
One photo becomes a personalized listing video — your face, your voice, your market — running automatically while you’re at showings. No more recycling the same script across the same platforms. No more wondering why the algorithm stops picking it up after two weeks. One photo is all it takes — one photo, and the AI puts you in the video, says your name, and speaks the right city every single time, automatically, while you sleep. You’re never re-uploading the same thing. You’re building a real brand, one unique video after another, on autopilot.
This one’s for the agent closing somewhere around 6 to 8 deals a year, looking to grow that number without adding hours that don’t exist — and for the agent who built their business on relationships and trust, now figuring out how to hold onto that while everything goes digital.
What It Actually Does
Here’s the part that actually matters, if you’re not a technical person, stick with us — this is in plain English.
You put one photo into the app, the photo is yours, you took it, you own the rights, and the app produces a video where you are the one talking, looking into the camera, introducing the property, saying the address, saying the number of bedrooms and baths, listing the features, closing with your name, your phone number, and your website.
The AI does the heavy lifting. It looks at the photo to figure out what kind of property it is, a house, an apartment, a commercial building, and it picks the right opening for the listing.
It knows what to call the different parts of the building — it says ‘kitchen’ not ‘room’, ‘bedroom’ not ‘space’, ‘bathroom’ not ‘room’.
It puts the right emphasis on the right words, so the finished recording sounds like a professional narration, not a script read aloud.
The consistency builds recognition — same face, same voice, same sign-off across every video, so your market starts knowing who you are.
How You Set It Up
Here’s the move-by-move build, because we know the technical version of this is sitting in front of you right now, and you’re thinking: I can’t code that. You can, in small steps, with Claude Code doing the heavy lifting each step of the way. If you’d rather watch us build it on screen, the free starter version of this, open-source, is on GitHub — just search ‘prosperity-real-estate-automation’. The version we run in our own business is the full automatic one, which runs on our machines unattended every day.
Step 1: Install the 3 tools you actually need. Claude Code handles the dependencies — you just open a chat and say ‘install MinIO, H3, and the video generation system on my machine’. MinIO is the cloud storage layer, H3 is the spatial index that lets the AI understand where a property actually is, and the video generation system is the part that actually makes the video — your face, your brand, on screen, every time. If you’re watching this on YouTube, the free starter version of that stack is in the description — you can run it on your own computer, right now, no charge.
Step 2: Point it at your own data. You run this once, from the command line, and tell it where your listings live — the file path, plain English, like ‘my listings live in /Users/alex/Documents/RE listings’. It connects to that folder, reads the structure, and from then on, every new listing you add shows up automatically. If you’re running the free starter version on your own machine, open the settings file and change the ‘data_directory’ line to point to wherever your listings live.
Step 3: Choose your voice. This is the engine that speaks your words, in your voice, on screen. The free starter version includes a high-quality, openly-licensed voice that you can use without asking anyone’s permission. If you want the highest possible quality — the same engine used in our own daily operations — you can upgrade to the full automatic version and run a single command to swap in your own, personally-licensed voice. Either way, the recording is done inside the system, in the cloud, so there’s no need for you to record anything yourself.
Step 4: Set the schedule. Tell the system when you want videos to go live — daily, weekly, whenever new listings come in — and it handles the rest automatically.
How It Gets You Found
Here’s the move that actually matters, and it’s the one thing nobody tells you: you are not competing with other agents online. You are competing with Google.
When somebody in your market types a question into Google — what’s a good neighborhood for families, how much does it cost to build a deck, is now a good time to sell — Google does not hand that person to an agent. Google hands that person to a page.
That page is either yours or it is somebody else’s, and there is no ‘agent website’ setting on Google. There is only ‘which page on the internet actually answers this question’.
Our system points every one of those questions, every single day, at a video on your own website that is already sitting in Google’s top 3, and it does that automatically, in the background, while you’re out at a showing.
You come back to the office, open the browser, and the report is already sitting there. Which query came in, what position it’s holding, and the exact wording Google put on the screen when your video showed up.
3 ways that work together to get you found, and the system sets all 3 up for you.
The first is the way we already talked about: the system takes the question, in plain English, and it uses that question as the exact file name of the video it generates. ‘Is it a good time to sell in Greenville’. One question, one video, one file name — and that is exactly what Google can understand and hand to a searcher.
The second is the way the videos are built to answer a question out loud, with a real person on screen, instead of just a talking head intro on a market update. When Google is trying to decide which page to give a searcher, it is looking for an answer that actually answers the question, clearly and directly — and this is that answer.
The third is the way the system publishes the video to the exact right page on your website, every time, instead of dropping it into a general ‘videos’ folder somewhere. The page it goes on is the one that’s already about that topic, the one that’s already getting some traffic — so the new video isn’t starting from zero. It’s building on what’s already there.
3 ways that work together, and you never have to think about any of them. You just tell the system what town you want to be found in, and it sets up the pipeline so every question that comes in from that market gets answered with one of your videos.
What It Means For You
For you specifically — the agent trying to stand out in a market where everyone’s photos look the same — this means the thing clients actually care about is finally within reach. It’s not a filter, a preset, or a piece of gear you had to buy. It’s your face, your voice, your explanation, on demand, every time a question comes up.
[CTA]
If you want to run this yourself — the free starter skills are in the description, and we offer a free 15-minute call to help you get started. Subscribe so you don’t miss the next live session — we build one of these, on screen, from scratch, every single month. We’re Al and Victoria Pinder, ICON agents at eXp Realty.
What It Frees You Up To Do
It’s compounding.
Every day you used to spend on a review that now takes 90 seconds adds up, and you’re the one deciding how to fill that time back in — more calls, more follow-up, more strategy, more bandwidth for the next deal.
If you’re scaling a team, it’s leverage. One system, multiple agents, consistent communication without anyone manually re-uploading a thing.
Either way, you built something you can walk away from, because you’re not the one inside it working a shift. You built a machine that works the shift for you, and you get to decide what that machine does next.
That’s what Prosperity means to us — owning your time, your image, your content, your audience, your brand, your clock, your day.
What To Do First
This is not about working harder. Harder gets you tired, not a business.
This is about the CEO move. Getting off the treadmill and letting the business work for you instead of the other way around.
The first instruction, and the one everybody resists, is: stop chasing leads.
Not ‘stop buying them,’ ‘stop paying for them.’ Stop chasing them.
Every platform has a hunger built into it — a hunger for more volume, more activity, more posts, more calls, more follow-up sequences running at once.
That hunger reads the same way to your business as it does to a prospecting newcomer’s. It reads as ‘needy.’
Needy is not how you get found by someone who has never met you. Needy is how you get ignored by someone who has been pitched 10 times this week already.
Here’s the counterintuitive move that actually works.
You do not need more leads. You need to be the one the lead goes to when it is ready.
Being the one the lead goes to when it is ready is not about posting more. It is about being the best answer the lead’s own phone can find, when the lead asks it a question out loud.
‘Siri, who is a good real estate agent near me.’
‘Google, what is a reasonable price for a house in my neighborhood.’
‘ChatGPT, how much does it cost to sell my house.’
Those are the moments you want to own, and you own them by being the best, most helpful, most trustworthy answer the assistant can give back.
Nobody types a question like that and then thinks ‘I should call the agent who posted the most this week.’
They ask the assistant, and the assistant gives them whoever it remembers, whoever it is told, whoever has earned the right to be included.
That earning is what this whole system is about. You stop renting attention and start owning it.
Before you build anything, you need to understand what you are actually asking the AI to do.
You are not asking it to make a video. You are asking it to take one photo and give you back a short video where you are the only person in it, looking into the camera, and the photo is on the screen behind you.
Where to start
If you want to run this on your own business — your face, your market — here’s your next step. We offer a free 15-minute call. It’s just us, Al and Victoria, and we’ll look at what you’re already doing and tell you honestly what fits and what doesn’t. The link to book it is right in the description below, and our number’s there too. And if you’d rather watch us first, our free, public live session is every Monday at 11 AM Eastern — that link’s in the description as well. We’ll put the free starter version of this, the one that runs on your own computer with no monthly charge, in the description too, so you can try it before you call us.
If you’d rather watch us build one of these, live, on screen, subscribe so you don’t miss the next session — the subscribe button’s right there on screen. We’re Al and Victoria Pinder, husband-wife ICON agents at eXp Realty, and we run this in our own business every single day. We’ll see you in the next one.
Frequently Asked Questions
How does the AI know what to say about each property?
The AI looks at the photo you upload to figure out what kind of property it is — a house, apartment, or commercial building — and picks the right opening. It also knows the correct terminology, saying “kitchen” instead of “room” and “bedroom” instead of “space,” then puts the right emphasis on words so the finished recording sounds like professional narration.
What tools do I need to install to set this up?
You need to install three tools: MinIO for cloud storage, H3 for spatial indexing that helps the AI understand property location, and the video generation system that creates your personalized videos. Claude Code handles the dependencies, so you can simply tell it to install these on your machine.
How does this help me get found by potential clients?
The system takes questions people search for and generates videos that answer them, then publishes those videos to the right pages on your website — pages that are already getting traffic and about that topic. This way your videos show up in Google results when someone in your market asks a real question, instead of competing with other agents posting the same recycled content.
A virtual assistant running your morning shift — answering client questions in your voice, pulling live data instead of canned answers, freeing you to close deals instead of typing recaps — costs less than you think and runs while you sleep. Here’s exactly how it works, layer by layer, because we run it in our own business every single day.
Who This Is For
If you’re a few years into this business, closing 6 to 8 deals a year, pushing for more volume and consistency — this is your play. You know the value of a fast response, a consistent brand, and a CEO-of-the-business mindset, and you’ve either built or bought some of it already. The question now is: what actually gets delegated to AI, and what stays human. Spoiler: the AI does the proving before you commit to anything.
What It Actually Does
Here’s the actual question, stripped of the marketing: can a chatbot replace a real assistant?
On a call, both will answer your question. The chatbot does it by searching its knowledge base — essentially, reading a manual written by a human, then parroting back whatever it found that matches the question.
The assistant reads the question, then does the actual work — she checks the listing, pulls the current status, and tells you what’s actually happening right now, not a generic answer written for anyone, anywhere.
On a slow week, that’s the same thing. On a busy one, it’s the difference between a client getting an answer today and getting a ‘I’ll look into it’ email that sits in a queue until tomorrow.
Nobody’s firing an assistant for typing fast. They’re firing her for not knowing something that was in an email she never saw, or for letting a question sit in a queue while the client went to a broker who answered it first.
The real question isn’t can a chatbot replace an assistant. It’s can a chatbot do the 80% of the job that’s clerical, so the assistant is freed up for the 20% that’s actual advice, actual advocacy, actual relationship work — the part that’s actually worth paying a professional for?
How You Set It Up
Here’s the honest setup, because we’ve run this on our own business. We’ve built two different AI help systems — one for quick, simple questions, and one for more complex, step-by-step guidance — and we’ve compared them side by side with the virtual assistant we used to pay. Same questions, same scenarios, filmed live so there are no edits hiding anything.
The simple Q&A system is set up with a few key intents — who we are, what we do, how to reach us — and a knowledge base pulled from our own website and FAQs. It answers things like ‘what areas do you cover?’ and ‘how does the buying process work?’ The same way, every time, because the information is pulled from one source and checked against it every time it’s asked.
The step-by-step guide system is the more interesting one, and it’s set up in Claude Code with a few Intents and a few Flows. The Intent picks up things like ‘what do I need to do before I buy a house’ or ‘how do I prepare my home to sell.’ Then the Flow takes over — it’s a set of defined steps, saved as a reusable pattern, with a starting point and a natural, conversationally-steered path through to the end. It doesn’t just recite a list — it asks questions, uses the answers to decide the next step, and logs the conversation so you can pick up where you left off later. And because the Flow is reusable, you change one pattern and the same structure applies to any matching question — you’re not rebuilding a separate answer every time something changes.
Both of these are running live, on our own business, while we film this. And the virtual assistant is right there next to them — same questions, same scenarios, so you can see the difference.
How It Gets You Found
Here’s the mistake we see agents make once they start automating anything: they assume ‘automatic’ means ‘get found.’
It doesn’t, not on its own.
An automation can generate a perfect blog post, a perfect social post, a perfect follow-up email — and that content can still sit unseen because the actual getting-found part is separate, and it’s not automatic at all.
Getting found is about two things, technically, and neither of them is about the content itself.
The first is the way the search engines read your site and decide what each page is about, so they can show it to the right person.
That’s done through something called on-page SEO, and while there is a right way to do it, there is no AI that consistently writes the right title tag, the right meta description, and the right alt text for the images without human review.
Those three things are what actually get you into the search results, and they have to be written by someone who understands them, either you or someone you hire.
The second thing is the way your content connects to your overall brand, so it’s clear to both people and search engines that all of this is coming from one real, real-world business.
That’s harder to describe technically, but it’s the reason why a blog post written in your voice, on your own site, is worth more than a guest post on somebody else’s platform — you’re building your brand directly, every time.
An AI can help with that by keeping your tone consistent, but it can’t replace the judgment call about what fits your brand and what doesn’t — that’s still you, looking at what you’ve published and deciding what belongs and what doesn’t.
Now, the review system we’re showing here does go further than just generating content.
It can look at what you’ve already published and tell you, honestly, which pages are performing and which ones are dead weight — that’s something an AI can answer that you probably can’t answer just by yourself, because it’s looking at your content as a whole rather than remembering a few popular posts.
That kind of review is what lets you stop pouring effort into the wrong things and put everything into the handful of channels that actually produce leads.
But even with that map, you’re the one who decides what to publish next — the review tells you what your audience wants, but you’re the one who picks the next topic, writes the headline, and sets the tone, because that’s the part that actually defines your brand.
What It Means For You
If you’re building a brand, this means you can answer a new client question instantly, consistently, and in your own voice — without handing that interaction to a third party to manage. If your business runs on relationships, every one of those relationships becomes sharper and more current, because nothing is lost in a handoff. And if you’re scaling, you get to choose exactly which queries get escalated to a human, instead of every single one landing in your inbox. You become the one who decides what’s actually urgent, instead of reacting to a pile of messages that all arrived at once because you were at a showing.
What It Frees You Up To Do
The real math here isn’t ‘can a chatbot answer this question as fast as a human.’ It’s what that hands back to you.
If you’re the one typing a recap after every call, or the one losing a night to a follow-up that wasn’t actually urgent, that’s hours you’re not putting into the next deal.
With this running in the background, you’re the one choosing where your time goes, instead of it being handed back to you in a stack of emails.
That’s real bandwidth, back for the owner who built their business on relationships and trust — now with the follow-through actually keeping up with the handshake.
What To Do First
Here’s the honest version, built for Google AND for the agent who’s trying to decide what to pay someone to do next.
If you’re a couple years into this business, you’ve probably already built or bought something like this — a shared Google Sheet with leads, notes, and follow-up items, or a Kluster board, or a Notion page, or all three, duct-taped together with a handful of automations.
Nobody tells you this before you build it: the real cost of that setup isn’t the few hours you spent on it. It’s that every time your business changes — you add a new market, or a new lead source, or a new team member — you’re back opening the keyboard, trying to remember where that column was or what that label was called, and either way you’re recreating the same argument you already had with yourself last time you touched it.
That is the whole case for this, in a nutshell. You stop building your own sources of truth, because they are always the first thing a change in your business finds friction with. You connect your tools to a real, running brain instead, and the friction becomes a question you ask out loud — what does ‘change’ even mean here, in my actual business, this week – and the answer comes back in the same language, useful for a decision, instead of a vague feeling sitting in a spreadsheet you can’t quite remember how to read.
Where to Start
If you want help building or upgrading your own AI tools, without the risk of building something that sits idle, book a free 15-minute call — the link is in the description. We’ll show you what we’ve built and tell you honestly what fits your business and what doesn’t.
The free starter skills are open-source on The Open-Source Claude Skills Repo for Real Estate Agents — that link’s in the description too. Clone it, try it, own it.
We’re Al and Victoria Pinder, husband-and-wife agents at eXp Realty, and every script, every schedule, every report runs in our own business every single day. We’ll tell you what’s real and what’s hype, because we’re the ones actually using it to run our business.
Frequently Asked Questions
What’s the difference between a chatbot and a virtual assistant?
A chatbot searches a knowledge base and parrots back generic answers, while an assistant reads the question and does the actual work — checking listings, pulling current status, and giving you what’s actually happening right now. On a busy week, that difference means a client gets an answer today instead of sitting in a queue until tomorrow.
How does the AI system actually get set up?
There are two systems: a simple Q&A system with key intents and a knowledge base pulled from your website for basic questions, and a step-by-step guide system built in Claude Code that asks questions, uses the answers to decide the next step, and logs conversations so you can pick up where you left off. Both are reusable patterns, so when something changes, you update one and it applies to all matching questions.
Will AI automation automatically get me found in search results?
No — automation generates content but doesn’t automatically get you found. You still need proper on-page SEO (title tags, meta descriptions, alt text) reviewed by someone who understands them, and your content needs to connect to your overall brand so both people and search engines see it as coming from one real business. The AI can help keep your tone consistent, but you decide what actually belongs in your brand.
One agent ran the numbers. The other didn’t. Same city, same property type, $500 difference in what investors actually offer.
The agent who ran the numbers owned the conversation. She knew the real yield. She knew the cap rate. She knew what the property was actually worth to an investor, not what the listing said it was worth to a buyer. When investors called, she had the answers. When competitors guessed, she had the math.
That’s the difference. An agent who runs live deal math on every listing doesn’t lose deals to agents who don’t know their own market.
If you are closing 6 to 8 deals a year and want to get the same kind of clarity before you take on a new listing, this is what we run for you.
If your business runs on relationships — repeat clients, referrals, the people already in your sphere — you know your biggest challenge isn’t getting leads, it’s holding onto the ones you have and showing them, in numbers, that you’re building real value under each property you take on.
Nobody expects you to be a financial analyst. But your clients do expect you to understand the real math behind an investment property, and whether a given deal actually stacks up.
If you’ve ever wanted to run those numbers yourself, without handing anything to a paid pro, this is your week. No spreadsheet background needed.
What It Actually Does
Here’s the actual sequence when you run this on a property. First, it loads every piece of relevant data about that property that it can find — the official listing, any associated MLS pages, the county’s own property record, and the local title company’s record, if that’s available where you work. Then it uses that combined picture to calculate five things, one after the next, in order.
Then it moves to the actual investment math. Second, it estimates what the property has actually earned for the owner so far, after holding costs — the difference between the rent it’s been bringing in and the expenses the owner has paid out, including mortgage, insurance, taxes, and whatever else is due every month. Third, it projects what that earnings pattern is likely to look like in the next six months, and the next 12, using the actual payment history the owner has built up so far, not just a guess about future rent. Fourth, it pulls the owner’s current mortgage balance and uses that, along with the projected rent, to calculate the actual return on that specific, real debt load — how much the property is earning the owner every year, relative to what they’re putting into it. And the fifth calculation is the one that actually connects all of the above to what an investor is actually trying to decide: it compares that real, debt-adjusted return to what a reasonable, diversified investment portfolio is likely to earn over the same period, using the established benchmarks for that.
That’s the whole point of the exercise. You’re not just showing an owner a pretty chart about their property — you’re putting that property into the same math an actual investor uses every day, side by side with the standard they’re already familiar with. And you’re doing it with their real numbers, pulled from their real life, not a hypothetical.
How You Set It Up
Here’s the actual setup, because we do these and you can build it yourself from this. We have 2 Claude Code skills running on a schedule, and you can find the exact instructions for both in the description — the free starter version is in the description too, and you can try it on your own machine before you put it on a schedule. The first skill runs every Sunday morning, and it does the heavy lifting. It connects to Realtor.com through their own API — that’s the real Realtor.com data, live, every week, not a saved copy from last month. It pulls every active listing in the county, filters them to the right criteria — ours is more than 2 acres, because we’re targeting land, but you can point it at anything: a specific town, a certain number of bedrooms, whatever your market needs — and then it calculates the 2 relevant metrics for each one, the way an investor actually looks at a deal. Number 1 is the gross yield, how much rent the property produces relative to its purchase price, and number 2 is the cap rate, how much cash it spits out in the first year relative to its purchase price, after debt service — both of those are common, practical, real-number metrics that actually get used in an investment decision. It writes both of those, along with the address and the listing link, into a plain Google Sheet — the sheet is the source of truth, and the video reads it live from the sheet every Monday. The second skill runs every Monday morning and turns that data into your marketing. It reads the latest row from the same Google Sheet — whatever is at the top — and uses that to generate the week’s email and the week’s social post. The email goes out Monday morning to the investor list, and the post gets scheduled for Tuesday morning, so the content is always fresh, it’s always about the current market, and you never have to touch it. If you’re thinking, I don’t have a list, start with the post. The scheduled post is what builds the list. Every piece of content points back to the same landing page, and that’s what the lead form lives on — so even if you’re starting from zero, you’re building an asset while you’re building the list.
How It Gets You Found
Here’s what actually gets you found, and it’s quiet: a Google-friendly page title that includes the town and the phrase ‘investment property,’ so you show up when someone’s actually searching for that, and a clear, repeated headline on the page itself, so both Google and the visitor understand exactly what the page is in about 100 milliseconds.
This is not about ranking for a keyword. This is about being findable when you are exactly what the searcher is asking Google.
The free starter skills are built so you can run this on your own machine, with no coding, and watch the page appear in front of you while you’re still on the call.
What It Means For You
If you’re building a real estate investment business, this means you can run a new deal across your own numbers instead of asking a broker or a partner for a read on it. If you’re pushing for volume, it means you can filter a list of 100 leads instead of calling them all blind. And if you’re running a small operation, it means you can build something like this once and reuse it, instead of rebuilding it from scratch every time a number changes.
What It Frees You Up To Do
The real math here isn’t about closing more deals — it’s about what you stop losing by owning the numbers yourself.
You stop losing hours to a third-party tool you don’t understand, bills you didn’t expect, and the quiet worry that the numbers you’re showing a client are someone else’s best guess instead of yours.
You also stop losing flexibility. If you’re building a real relationship with a real investor, you know the moment a set standard answers stop being the right answer — the moment a deal is actually worth walking away from a percentage split.
Nobody outside this room can tell you what that moment is worth to your business.
If you’re building a business where every deal matters, every relationship matters, every call matters — this is what lets you go into every one of them knowing exactly what you’re looking at.
What To Do First
If you’re pushing for volume right now, go make a call instead — the pipeline is more valuable than this exercise.
Not ‘build a website’ or ‘post on social.’ Those are activities. This is an assignment: run the numbers on what it actually costs to acquire a deal, what that deal actually generates, and whether the gap between the two is getting bigger or smaller.
You do not need to be a programmer. You open Claude Code, you type in plain English — ‘run the acquisition cost calculation,’ ‘show me the ROI on the last 6 deals,’ ‘what’s the gap between what we’re paying to get a deal and what it actually returns’ — and it builds the actual calculation and shows you the real math, live, on the screen, without you writing a line of code.
The version of this one that’s already built is called ‘deal-economics.’ Open source, free, no sign-up. The repository link is in the description, and the whole thing runs in the free tier — no cost, no trial, just install it and run it.
If you want to run this on your own business — not just watch us do it on YouTube — here’s your next step.
We offer a free 15-minute call. It’s just us, Al and Victoria, and we’ll look at what you’re already doing and tell you honestly what fits and what doesn’t. The link to book it is right in the description below, and our number’s there too.
Or start tinkering yourself. The free starter skills are open-source on GitHub — that link’s in the description as well. Clone it, try it, own it.
We’re Al and Victoria Pinder, husband-and-wife ICON agents at eXp Realty — and we run this in our own business every single day.
What math does this tool calculate for investment properties?
The tool calculates five things: it loads all relevant property data from listings and county records, estimates what the property has earned the owner after holding costs, projects earnings for the next six to twelve months based on actual payment history, calculates the real return on the owner’s specific debt load, and compares that return to what a diversified investment portfolio typically earns using established benchmarks.
How often does the system update with new listings?
The first skill runs every Sunday morning and pulls active listings from Realtor.com’s API, then writes the calculated metrics to a Google Sheet. The second skill runs Monday morning, reads the latest data from that sheet, and generates fresh email and social content based on the current market listings.
Do I need coding experience to set this up?
No coding experience is required. You can open Claude Code and type plain English requests like “run the acquisition cost calculation” or “show me the ROI on the last 6 deals,” and it builds the actual calculations and shows you the results live on screen. The free starter skills are also available open-source on GitHub and run on the free tier with no cost or sign-up.
Three patterns are working right now for agents building consistent volume: a question-and-answer system that answers listing questions in seconds, a review system that audits what’s actually working in your content, and a draft tool that handles the repetitive writing. Here’s what each one actually does, how to set it up, and what it gives you back.
Who This Is For
If you’re a few years into this business, closing 6 to 8 deals a year, pushing for more volume and consistency — this is your play. You didn’t get into real estate to become a part-time IT specialist, and you don’t need to build anything from scratch to get the compounding benefits of AI on your business. This is also for you if you’ve already tried a tool or two and they felt more like a toy than an actual help. We’re going to show you what real AI assistance looks like, running behind the scenes, working the same hours you do.
What It Actually Does
Here’s the mistake we see agents make: they ask Google Gemini, or ChatGPT, or Gemini, ‘write a real estate blog post’ and call it a day. That’s not any of these 3 actually doing the work — it’s you, typing out a question, and then reading the answer back to yourself. The AI didn’t write the post.
The 3 patterns that actually work, across all 3 tools, are these. Here’s Pattern 1: a question-and-answer system — you ask it a question about a listing, or a market trend, or a client question, and it gives you an answer, drawn from sources it can actually verify, in a few seconds. No typing out a full sentence — you just ask it, the same way you’d ask a team member. Pattern 2 is a review system — you give it a piece of content you’ve already made, a video or a post, and it tells you what’s actually working and what’s just filling space, so you know what to keep and what to cut before you publish. Which headline is actually a question people search, and which is just a statement you liked? Which section is actually backing up the lead with a source, and which is just an opinion thrown in? Pattern 3 is the one people picture when they think of AI writing: you set a topic, a tone, and a length, and it writes the draft. Then you review it, ask it to change something if you want, and once you’re happy with the draft, you publish it. The draft is yours to own — you’re the one deciding what goes in, what stays out, and what becomes the headline. None of these 3 patterns replace you with a bot. They replace typing, scrolling, and second-guessing with asking a question and getting an answer. You’re still the agent — now you have 3 tools that work the same way, across different jobs, and you pick the one that fits the task.
How You Set It Up
Here’s the honest version, because we’re all engineers at heart. None of this is the ceiling on what an AI can answer — it’s just what’s been built so far, and the manual work doesn’t go away just because you’re paying for a seat at a webinar. If you want to run any of this yourself, the free starter skills are open-source on The Open-Source Claude Skills Repo for Real Estate Agents, and the version of Claude Code we used is the free tier, with the receipt in the description. Everything else is a browser tab and a credit card. The manual part is putting it all on one schedule, and that’s the part that actually makes you more efficient — not any single tool.
How It Gets You Found
Here’s the mistake we see agents make once they start exploring AI: they ask Google something like, what’s the best AI for real estate, and they take whatever comes back at face value. None of that tells you anything about what’s actually working in the real world, right now, for real agents doing real deals.
So we did the opposite of asking Google. We went straight to the source.
We reached out to the agents who have already built AI into their business — the ones who are actually using it every week, not just the ones who wrote a blog post about it once. We asked them what’s working, what’s a waste of time, what’s worth paying for, and what you can build yourself with a little technical help. And we did that before we built anything, because we didn’t want to spend months developing something nobody actually needs.
What we found is that there is no one ‘best’ AI for real estate — there are 3 different jobs that AI does for an agent, and 3 different tools built for those jobs, and the right one for you depends on what you’re trying to achieve. Some of these tools are brand-new. One of them didn’t exist 6 months ago. They are not the same tool, they do not compete with each other, and there is no marketing hype behind any of them. You are not going to find any of this by searching Google.
The first job is the one everybody thinks of: getting found. People search Google, they ask ChatGPT, they type a question into their phone — how do you get the answer that comes back to be you, the local agent, instead of a big national site or a generic answer? Nobody types ‘hire a real estate agent’ anymore — they ask a question, they describe a situation, and the AI answers it. Being findable inside those answers is the new front door to your business, and it is not the same as ranking a website.
What It Means For You
If you’re pushing for volume, this turns a 40-minute evening job into a 2-minute review. If you run your business on relationships, every checked-in property is a record of a real touch, and nothing goes lost into a notebook. And if you’re building a brand, this is how you get consistently good-looking content out, every week, without it eating your whole Sunday.
What It Frees You Up To Do
In the end, none of this is about the AI itself — it’s about what it gives back to you. You build your brand on your own terms, instead of handing it over to a platform that rents it back to you. You walk into every call already knowing more about the client than they expected, without spending an hour on a questionnaire. You stop paying a marketing agency to do what you can now run yourself, across every listing, every week, with a consistent look that can’t be duplicated by anyone who didn’t build it. And you do it all with the same 24 hours you’ve always had — you just stopped wasting them on the things that now work quietly in the background, without you touching them.
That’s not a shortcut to the same old business, and it’s not about doing more. It’s about building something you actually own, layer by layer, with the client standing in front of you as the real audience every time.
What To Do First
If you’re building a real estate brand online, the best place to start is almost never ‘post more.’ It’s almost always: what is your audience asking, and is there a video or a page already out there answering it? Nobody types a question into Google and adds your name. They ask Google, or ChatGPT, or their phone: what’s a good neighborhood for families in your city, how does the buying process work, is now a good time to sell. Those are the exact same questions inside Google’s language models, the exact same suggestions auto-complete is putting in front of people every single day. Being findable means matching the way your future clients already search, and matching the answers the AI tools are already giving them — because those answers are what set the standard for what a good answer looks like, whether you’re a human or a bot writing it.
Here’s the part nobody tells you: you don’t have to write the perfect answer on your own. You ask ChatGPT, or Gemini, or Perplexity: what does a buyer need to know before they start looking at homes in my market, and it gives you a list. You are not inventing anything. You are taking the actual questions your future clients are already asking, and giving them the best version you can, in your voice, on your website, so you’re the one showing up when someone from your area types that into Google next Tuesday.
The mistake is thinking you need a piece of software to do that. You don’t. You need a list, and a schedule, and the willingness to sound like a real person writing to your neighbors instead of a marketer writing to a target demographic. That is the whole game — being findable by the people who are already asking, and sounding like someone they want to talk to when they do.
Where to Start
We built this system running our own deals, so we know what actually works in the field. If you want help picking the right AI tool for your own business — the right one for you specifically, not a generic list — we do a free strategy call. It’s just us, looking at what you’re already doing and telling you honestly what fits and what doesn’t. No pitch, no obligation, just two agents who run AI in our own business every single day telling you honestly what fits and what doesn’t. The link to book is in the description below, and our number’s there too.
Subscribe so you don’t miss the next session. We’re Victoria and Al Pinder, ICON Agents at eXp Realty.
Frequently Asked Questions
What are the 3 patterns of AI that actually work for real estate agents?
Pattern 1 is a question-and-answer system that answers listing questions, market trends, or client questions in seconds using verified sources. Pattern 2 is a review system that audits your existing content to show what’s working and what’s just filling space. Pattern 3 is a draft tool where you set a topic, tone, and length, then review and edit before publishing.
How do I set up these AI tools for my real estate business?
The free starter skills are open-source on The Open-Source Claude Skills Repo for Real Estate Agents, and you can use the free tier of Claude Code. The key manual work is putting all three patterns on one schedule, which is what actually makes you more efficient—not any single tool by itself.
How does using AI help agents get found by potential clients?
People search Google, ChatGPT, or their phones with specific questions about buying, selling, or neighborhoods. By answering the exact questions your future clients are already asking—in your voice and on your website—you become the one showing up when someone from your area searches those questions. This matches how AI tools already answer those questions and sets the standard for what a good answer looks like.
If you’re pushing past 6, 7, 8 deals a year and you’re still doing that kind of research one manual search at a time, it’s not that you’re behind — it’s that you’ve outgrown the free tools, and you haven’t been shown what’s actually available next.
You don’t need a tech background. You just need to be the one who says yes when someone asks, ‘Who does all the behind-the-scenes stuff for your business?’ Because that person is you, whether you built it or bought it. And either way, it’s sitting in the room while you watch this, waiting for the next question.
What It Actually Does
Here’s the swarm, actually running. It starts from a clean machine — just the free, downloadable skills we listed earlier, running on a Mac. No coding, no monthly seat, no API key we paid for. The first skill calls out loud, into the room, ‘what’s the address of the property we’re checking today’. We tell it the address, out loud, and from there the whole swarm runs on our ice. It’s the same idea as the old cliché, asking Siri or Google a question — except this question isn’t routed through a company’s help desk, and the answer isn’t a canned reply written by a marketing team. It’s a real query, directed at real data, fired the same way every time, so every answer is comparable to every other answer, without us touching anything technical.
The second skill pulls the property’s own listing — what’s actually posted about this house, right now, by whoever posted it — and reads it back to us out loud, so we’re looking at the same thing the client’s looking at, without needing to log into anything. That instruction’s the same every time, too: ‘read the listing for this address’. The swarm learns the routine, and every property becomes just a matter of repeating the same few phrases.
The third skill is the one that actually does the digging. It goes after six specific signals, one after the other, and writes them down into a single, formatted report, which it saves as a file we can open later, side by side with next week’s report, or next month’s. The first signal is the simplest: does anything about this property show up in Google News, anywhere in the world, in the last 90 days? That’s not a Google search — that’s a directed query against Google’s own news index, the same way a journalist would query it. It’s the signal that matters most when a client’s asking something like ‘has there been anything published about this place that I should know about’.
The second signal looks outside Google, to the enthusiast sites — the real estate blogs, the local news portals, the forums where people actually talk about properties, instead of just posting them. That’s a different set of sources entirely, targeted specifically for the kind of local, ground-level conversation that filters up from a real community, rather than a press release handed out to wire services.
How You Set It Up
Here’s the actual swarm running. It’s just a Google Sheet — no coding, no API keys to renew. Every morning at 8:05, it reads yesterday’s date into the top. Then each agent’s row runs itself, one after the next, and you can watch it work in real time.
The first column is the check. It runs a quick sanity check on the whole row before it does anything else. If the landing page URL is missing or broken, or the Google Alerts email address is wrong, it catches it and skips that run instead of quietly failing later. That’s the whole point of the swarm — you’re not watching one job crash, you’re watching 6 or 8 or 10 independent runs, and if one fails it fails quietly while the others keep going.
It does 3 things, in order, every morning.
First, it pulls your Google Alerts email for the brand keywords you set up last week — your name, your team, your town, your website. Not a Google Search — a real, live email digest, delivered overnight, checked fresh every morning. That’s the part that actually matters. A Google Search window-shopped at 9:15 AM shows you what was popular yesterday afternoon — a brand alert checked at 8:10 AM shows you exactly what was published about you, your team, or your market while you were asleep.
Second, it pulls your Facebook Page metrics for the last 7 days — reach, impressions, post engagement, video views, whatever matters to you. Not a screenshot, a real download, saved as a CSV you can open and filter later.
Third, it pulls your YouTube channel metrics for the last 28 days — views, watch time, impressions, clicks, whatever matters to you. Same as Facebook — a real download, not a screenshot.
Those 3 things cover the 3 ways a local brand actually gets found today — Google searching, social scrolling, and video watching. And they cover the 3 sources you actually own. You own your Google Alerts setup. You own your Facebook Page. And you own your YouTube channel — the content, the brand, the audience, the data, all of it.
How It Gets You Found
Here’s what actually gets you found, scene by scene, and it’s built into the swarm from the very first run.
The first thing is the query target itself. Every one of these is a real question, a real phrase, a real way someone searches. ‘Homes for sale in [city]’ is the exact target. ‘Homes for sale near me’ is the exact target, with the added benefit of picking up the local version of that search in Google. ‘Real estate agents in [city]’ is the exact target, with the added benefit of owning that whole map pack. And ‘how much does it cost to sell my home’ is the exact target, with the added benefit of owning that whole featured snippet space. Those are the exact searches, the exact targets, that the whole swarm is built to serve.
The second thing is the way the content gets built. Every piece of content becomes a standalone, indexable page, published on your own site, under your own domain, your own brand. That’s the actual real estate asset that’s left behind, long after the swarm has finished running. It’s not a post, it’s not a piece of guest content, it’s a real, ownable page on your site, built from the actual questions your future clients are asking.
The third thing is the way the images get used. Every piece of content includes at least one relevant, correctly sized, correctly named image. ‘Relevant’ means it matches the topic, the query, the search. ‘Correctly sized’ means it’s the right aspect ratio for the job, either a vertical image for a vertical format or a square image for a square format, because the AI tools will automatically resize a correctly sized image to whatever ratio the output needs, but they can’t fix a mismatched aspect ratio. And ‘correctly named’ means the file name includes the target keyword, so the image itself is something a search engine can actually understand and index. Together, those three things mean the images themselves become searchable assets, indexable by Google, and that indexing gets compounded every time a piece reuses that image on a new piece of content, which is exactly what the swarm is designed to do.
The fourth thing is the way the URLs get written. Every piece of content gets a URL that is, in order of importance, short, descriptive, and accurate. ‘Short’ means it’s not a 190-character random string — something like `/homes-for-sale-raleigh` instead of `/property-search-results-filter-bedrooms-price-sq-ft-year-built-hoa`.
What It Means For You
If you’re pushing for volume, this is how you stop each listing getting 3 hours of your evening that add up to 0 value. If your business runs on relationships, this is how you stop your assistant handing your brand to someone you’ve never met. And if you run on a niche, this is how you stop a generalist swarm handing your niche back to Google. Same system, sourced the same way, now you choose who touches it and why.
Both free versions are in the description — the swarm one, and the version you run yourself on your own machine without needing any API keys or logins. If you want help setting up either one, or you’re building something bigger and you’re not sure where to start, book a free 15-minute call, link’s in the description, and we’ll show you what it can do running live. Subscribe so you don’t miss the next live session — we build one of these, on screen, from scratch, every month. We’re Al and Victoria Pinder, ICON Agents at eXp Realty.
What It Frees You Up To Do
So what actually gets handed back to you? That’s what you physically receive — but the real math is what comes next.
Every hour you used to spend preparing for a client call is now an hour you can put back into the market, into the next call, into the next follow-up. You’re effectively adding hours to your day, and you’re adding them in the exact place where every agent wants more bandwidth — on the front end, before the client has even signed.
Think about what that actually does for a business that runs on referrals. Every satisfied client becomes a stronger referral source, because the prep work that used to let a call go stale now guarantees that call is sharp the moment it’s scheduled. And for an agent building a niche brand, every targeted call becomes a sharper piece of content, because the prep work is already done — it’s the exact same swarm run you just used on a real client, spun out as a blog post or a video script with a few modest tweaks.
Either way, you’re the one who decided what runs and what sits. The swarm is the engine, but you’re the pilot. And that, honestly, is what makes us most excited about where this is heading. We didn’t build any of this to watch from the sidelines — we built it so that the next agent to run something like this on their own machine is watching a video we made while they’re doing it.
What To Do First
If you’re like most agents we talk to, you’ve heard ‘swarm’ or ‘AI’ thrown around and thought: that sounds like something I should be doing, but I don’t have time to learn to code, and I’m not technical. And you’re not alone — the vast majority of agents who contact us describe themselves as non-technical, and that’s exactly why we built this. You don’t need to understand how any of it works, you just run it from the command line, the same way you’d open a Google Doc. It’s the exact same principle as how you already use a camera — you don’t need to understand the mechanics of the lens to take a good photo, you just point and press the button. This is the same idea, applied to research. You point it at a market, a neighborhood, or a specific property, and it runs the swarm for you — no coding, no learning curve, no monthly fee. You own it, you run it, on your own machine, any time you want.
Where To Start
If you want to run this on your own business — not just watch us do it on YouTube — we’d love to talk. We offer a free 15-minute strategy call. The link to book it is in the description, right below the one to the free version of Kimi on GitHub.
On the call, we’ll show you exactly how this works with your own data, not a demo. And we’ll tell you honestly what fits your market and what doesn’t.
We’re Al and Victoria Pinder, husband-and-wife ICON agents at eXp Realty here in Eastern North Carolina, and every script, every schedule, and every report runs in our own business every single day. If you’re building a brand, a niche, or a business that runs without you, this is the next piece that puts it all on a schedule.
Frequently Asked Questions
What skills does the swarm use to research a property?
The swarm uses three main skills: the first asks for the property address out loud, the second pulls and reads back the property’s current listing, and the third digs into six specific signals including Google News results from the last 90 days and mentions on enthusiast sites like real estate blogs and local forums.
Do I need coding knowledge or technical skills to run this?
No, you don’t need any coding background or technical skills. You run it from the command line the same way you’d open a Google Doc — you just point it at a property, neighborhood, or market and press the button. It works on a clean Mac using only free, downloadable skills with no API keys or monthly fees required.
How does the Google Sheet swarm help my business each morning?
Every morning at 8:05, the swarm automatically pulls your Google Alerts email for brand keywords, your Facebook Page metrics for the last 7 days, and your YouTube channel metrics for the last 28 days — all saved as real files you can open and filter later. This covers the three ways your local brand actually gets found today: Google searching, social scrolling, and video watching.
We run 3 different AI tools every morning before we open a single app. One checks what’s actually trending in Greenville real estate that week, not what trended last month. The second reads through the day’s new listings and pinpoints which ones are actually worth a call, without us reading a single description. The third watches our own content across all 4 of our brands and tells us exactly which videos and pages are getting found by Google and which ones quietly stopped mattering. None of that is us sitting down and ‘keeping up with things’ — it’s a machine reading 10 times more sources than we ever could, and we just review the digest. That’s the real gap AI closes for a busy agent: not doing our job, but doing the part of our job that used to eat an entire morning.
Related: Claude Code for Real Estate Agents — Our Stack
If you’re a few years into this business, closing 6 to 8 deals a year, pushing for more volume and consistency — this is your play. You didn’t get into real estate to become a marketing expert, but you also know being invisible online means losing deals to the agent who shows up in a Google search. This is how you stop feeding the portals and start building something that works for you.
What It Actually Does
That’s not what any of these tools actually are — they’re not a separate, automated version of you. They’re a smart assistant, sitting next to you, doing the heavy lifting on tasks you’ve already been doing with your own brain.
Closer look, tool by tool.
ChatGPT and Google Gemini both answer questions and generate text, and they do it in slightly different ways. ChatGPT builds on the conversation history — ask it about a property, then ask a follow-up, and the second question gets answered using the first one as a reference.
Which one is right for you depends on the job. If you’re drafting a letter where the context matters — a reply to a specific client question, for example — ChatGPT’s conversation-based approach keeps that context alive the whole time. If you’re generating a batch of descriptions where each one stands alone, Gemini’s separate-question approach means you’re not accidentally building up a shared ‘memory’ between one listing and the next.
Perry Gemelli’s AI tools are a different animal — they’re built for agents who want to run something on their own machine, without a monthly subscription. The free starter version runs on your own computer, costs nothing, and there’s a version you can buy that runs faster and handles more listings at once. If you’ve ever opened up a skills plugin inside Claude Code and thought, I could build that, this is what that build looks like — the code is open, you can read it, and you can put it on your own machine tomorrow if you want to.
The listing-description tools are the ones that get the most attention, but they’re actually the smallest piece of what’s actually possible here.
That preview feature is the one that actually saves real time — not drafting faster, but catching a wrong field or a missing image before anyone else sees it.
How You Set It Up
Here’s the actual setup, because we want you to walk away with something you can actually build. We’re not showing you a finished, polished website — we’re showing you the underlying engine, running live on screen, while we explain it. And we’re doing it piece by piece, so you can see exactly what’s doing the work, and why it’s arranged that way instead of some other way.
The first thing that runs every single morning, before you’ve even opened your laptop, is the content brief. It’s set up as a cloud function — one scheduled job that fires a single command, from the computer, to the cloud, and then goes back to sleep. When you’re the one who built it, ‘scheduling a cloud function’ just means opening one settings file and changing a number. It’s the same thing as setting an alarm on your phone, except the alarm is the computer, and the wake-up call is a perfectly written blog topic, already sitting in your inbox, ready to be typed out.
Here’s the part that actually matters: the brief pulls real questions from your market, not random topics from a keyword list. It’s a real question, phrased in plain English, like ‘what’s the difference between a condo and a townhouse’ or ‘how does the home buying process work for first-time buyers’. Those are actual examples from last week. And here’s the part that makes it useful: the question is always about something you can actually answer, with your own experience, your own knowledge of your market, built into the answer. You are the expert the AI is asking, and the answer comes back formatted as a real question-and-answer post — the question at the top, the answer below it, both in a shape that’s easy to read on screen and easy for Google to understand. So the post is useful to the reader before you’ve even opened it, and the target keyword is already in the title — not suggested later, not typed in afterward, but built into the question before you’ve touched it.
Now here’s the part that usually surprises people. The post is not a draft sitting in a needs-approval queue. It’s a real, published post, sitting on your website, dated for today. The schedule is built into the one command you run every morning. Some days it publishes a market update. Some days it publishes a FAQ. Some days it publishes a neighborhood spotlight.
How It Gets You Found
Here’s the mistake we see agents make: they ask ChatGPT, ‘What’s a good real estate headline?’ and it writes something generic that 500 other agents also asked it to write. Same with the descriptions — vague, unremarkable phrases that the AI hands out by the thousand because it doesn’t actually understand what makes a listing show up.
None of that is the same as being found.
Being found is walking past a house and the Google Autocomplete finishing your sentence before you’ve typed it out, because Google already knows that street is a listing. Being found is the AI tool you hired to watch your market finding a brand-new listing before anyone else has typed its address into anything, because the tool reads the source material the human never has time to open.
Being found is owning a niche so completely that when someone in that niche asks Google a question, you’re the one who shows up and answer it — because you’ve built the website, you’ve written the content, and you’ve told Google in multiple ways exactly who you are and who you serve.
None of that is a headline template. It’s a map to how the AI tools can work the same hours you already work, watching your market while you’re at a showing, checking sources you’d never have time to check yourself, and handing you the proof the next morning.
The old way of getting found was publishing more than everyone else. The new way is owning a niche so completely that the AI tools can confirm it for you every single day.
If you want to see exactly what that looks like — the map, the tools, the owning — book a free 15-minute call with us. The link is in the description, and we’ve dropped a free starter version of the system into the description too, so you can see the first stage of it running before you ever talk to us.
What It Means For You
For you specifically — the agent trying to choose what’s worth the effort — this means you can stop building your own AI tools and stop paying someone else to build them for you. Pick the 2 or 3 jobs you actually do every week, and the matching skill is already out there, free, ready to run on your own machine, on your own terms. You own it. You understand it. You decide when it changes.
If you’re pushing for volume, repeatable checklists become reusable code. If you’re building a brand, consistent content becomes a pipeline you feed, not a post you write one by one. And if you’re protecting a transaction, an alert system is what stands between you and a surprised client.
None of this is the ceiling on what AI can do for you. It’s just what it’s actually asking, right now, in plain English, with real examples. If you want to see what else is out there — what’s coming next, and what’s still just a proof of concept — we offer a free call where we can show you more, live, and tell you honestly what’s worth watching versus what’s a demo in search of a problem. That link’s in the description. Subscribe so you don’t miss the next live session — we build a full system from scratch every month.
What It Frees You Up To Do
The real math here isn’t ‘I typed a question and got an answer.’ It’s compounding.
Every evening check you’ve now got back is an evening you can spend on the next client, the next call, the next follow-up, instead of re-typing what you already have.
That compounds into more pipeline, more closings, more volume, without adding hours that don’t exist.
If you’re building a business, not just a paycheck, that’s the real ROI.
If your business is already running smooth and you’re looking to actually grow volume, the real payoff is the consistency.
The consistent brand messaging, the consistent follow-up, the consistent content — that’s what gets you found, gets you referred, and gets you remembered, and this is what finally lets you run that pipeline in parallel instead of one deal at a time.
You become the obvious choice in your market instead of the hard-working agent who’s always too busy to follow up.
Either way, it’s the same 2 concepts, applied to whatever you actually sell.
You own the infrastructure — the data, the brand, the relationships — and you license the access to it, on your terms, for the specific client or the specific deal.
That’s what makes you the one calling the shot, instead of the one asking permission.
If you want to see exactly how we run this in our own business, the free starter skills are open-source on GitHub — that link is in the description, along with a free 15-minute call if you’d rather talk it through. Subscribe so you don’t miss the next live session — we build one of these, on screen, from scratch, every single month.
What To Do First
If you’re building a real estate brand online, you’ve probably already got a website — and you’ve probably heard someone tell you to ‘just start blogging.’ Everyone says that, but here’s the part they don’t tell you: a blog post is not the same thing as a lead. A blog post is a piece of content you publish and hope gets found. A lead is a real person, with a real question, who lands in your inbox because you answered that question exactly right — and today, answering that question exactly right almost always means using AI, in one of these three ways, which are the three things we do first with every agent who partners with us.
Where to start
If you want to run these on YOUR business — not just watch us run them on ours — here’s your next step. We offer a free 15-minute call. It’s just us, Al and Victoria, and we’ll look at what you’re already doing and tell you honestly what fits and what doesn’t.
The link to book it is in the description below, right next to the free starter skills you can download and run yourself. And if you’d rather just watch more, subscribe so you don’t miss the next live session — we build these on screen every month.
Frequently Asked Questions
What’s the difference between ChatGPT and Google Gemini for real estate?
ChatGPT builds on conversation history, so follow-up questions use previous context as a reference—useful for drafting client replies where context matters. Google Gemini treats each question separately, which works better when generating multiple listings where you don’t want shared ‘memory’ between them.
How often do the AI tools run, and what do they check?
The tools run every morning before you open any apps. One checks what’s actually trending in your local real estate market that week, another reads new listings to pinpoint which ones are worth calling on, and the third watches your content across all your brands to see which videos and pages are getting found by Google.
Do I need to pay monthly for these tools?
Perry Gemelli’s AI tools are built to run on your own machine without a monthly subscription—the free starter version costs nothing. There’s also a paid version that runs faster and handles more listings at once, but you own the infrastructure and can decide whether to upgrade on your terms.
Say you listed a $300,000 home in Lumberton on Monday. By Wednesday, two new comps hit the market three blocks away. Your CMA from Monday is already stale. Do you pull a new one? Wait for Friday’s data? Or do you walk into a follow-up call knowing the system ran the numbers overnight and your current comp set is sitting in your inbox, dated today, before you even had to think about it? That’s what this system does. It doesn’t wait for you to remember. It doesn’t care if it’s Wednesday or Sunday. It just runs the comparison every single time something changes, and you get the fresh answer automatically.
Related: Claude Code for Real Estate Agents — Our Stack
If you’re a few years into this business, closing deals and building momentum, listing copy is one of those tasks that eats an evening every time — you type, you second-guess, you wonder if a phrase you’ve used 100 times is actually a compliance problem. And it’s the exact same doubt, repeated, for every single listing. Same with the MLS input. Same with the follow-up sequence you know you need to build, that you keep putting off because it’s the least glamorous thing in the world to sit down and code. None of it is the CEO move. None of it is why you got into this business. But all of it is exactly the kind of repeatable, predictable work that AI can own, running in the background, without you touching it.
What It Actually Does
Here’s what each one actually does, in plain English. The first three — the comp search, the market overview, and the neighborhood description — are the ones most people think of when they think of an AI doing a CMA. The comp search is the biggest: it’s not just looking at what’s sold recently, it’s also filtering out listings that are obviously not relevant, and it’s doing that across multiple sources, not just one. That part alone is worth it, because a human version of that task takes real time, and it’s easy to miss something when you’re tired.
The market overview and the neighborhood description are the other half of why an AI can actually give a client something a human can’t: a fresh, current read of the market, built every single time, without anyone remembering to go update a template. That’s what lets you walk into a follow-up call knowing you have the absolute latest numbers in front of you, instead of reading last month’s market update and hoping nothing changed.
The other two are the ones we think are the most interesting, because they’re the ones most people assume can’t be automated. The first is the manual read-through of the actual MLS description, checking whether the square footage matches the listing and whether the property details actually match what’s shown in the photos. That’s not a question of how many sources it searches — it’s a question of whether the system can actually read the description, parse the numbers, and compare them against the pictures, which is a step most AI tools skip. Ours doesn’t skip it, and that matters because a listing with the right label but the wrong square footage in the description is exactly the kind of thing a human checker catches — and an unchecked mistake like that is exactly the kind of thing that leads to a client asking, ‘How did you get the facts wrong on my home?’
The second is the part almost nobody expects. When you’re trying to get a read on a specific neighborhood, you don’t just look at what’s sold there — you look at what’s about to sell. A neighborhood with 5 deals pending in the next 30 days is heading in a different direction than one where the last closing was 3 months ago, and that read gets lost the moment you filter for only ‘closed’ sales and miss what’s about to close. Nobody types that filter on purpose — they type ‘neighborhood X’ and get whatever came up last time, which may or may not be current.
How You Set It Up
Here’s the honest version, because we’re not trying to talk you into anything. We run this on our own business every single day, and we’re happy to show you exactly how we set it up. The free, open-source version of this, the one you can run on your own machine right now, is on GitHub — that link is in the description, and the name is Claude Code CMA, and it’s been used by agents all over the country, not just us.
Now, the plain-English version, because we’re agents first and technologists second. A common question we get is, what does ‘own’ actually mean here? It doesn’t mean the AI is typing out an email and hitting send — that’s a compliance nightmare, and we wouldn’t recommend it to anyone.
What it actually means is that the output becomes a scheduled job, the same way a recurring alarm on your phone is a job that fires automatically on a schedule. Every morning, before you’ve even opened your laptop, this system checks for new listings in your market that match your client’s criteria, and if it finds any, it drafts a digest in your voice and sets it to go out later that day — the timing is your call, you just tell it when you want the digest to land.
It’s the same principle as a smart watch checking your heart rate every morning while you’re still in bed — you didn’t have to remember to open an app and run a report, the check happened automatically on a schedule, and if something was off, you’d have the digest sitting in your inbox telling you exactly what changed.
Now, the part we want to be really clear about, because it’s the piece that matters most if you’re thinking about your own business — you are never locked into anything. The free, open-source version runs on your own machine, on a schedule you set, on a job you can cancel any time. If you decide to run it through a subscription service instead, you’re not signing a contract — you subscribe month to month, the same way you would with any other tool, and you can cancel any time.
How It Gets You Found
Here’s the actual question Google gets asked, over and over, by real people in your market, your language, your exact words: what’s a home worth near me. That’s a question with a location in it, a question Google understands as I want to know something about where I am, and Google answers it by looking for a local, published, recent answer — a recent post, a recent page, a recent report, something it can hand to the person who asked and say here is your answer. That is exactly what this system produces, every single time, automatically, without you touching it: a published, date-stamped, location-specific answer to the question people in your market are actually asking Google. And there are plenty of other places that question gets asked now besides Google — inside ChatGPT, inside Gemini, inside Perplexity, inside Claude — those are the new answer engines, and this system gets you quoted in the answers they return to your market, automatically, every single time. You become the source the AI tools quote when somebody asks about your market, your listing, your name right there in the answer, over and over, automatically, while you sleep, while you’re at a showing, while you’re closing. That is real traffic, real exposure, real credibility, built once and running forever on a schedule, on a system you own, on a machine that’s already paid for.
What It Means For You
That’s real bandwidth back — bandwidth you can put into the next client call, the next follow-up, the next relationship, instead of into an administrative task that’s now running automatically behind the scenes.
What It Frees You Up To Do
The real math here isn’t about saving a few minutes — it’s about what those minutes add up to over a month, or a quarter, or a year. That’s real bandwidth, back to you, to put into what actually grows your business.
If you’re pushing past 6 or 8 deals a year and you’re honestly trying to scale that number, look at where your time actually goes, and ask yourself which of those hours is actually the best use of your energy — the CEO energy, the relationship energy, the sharp-edge energy that can’t be handed off. Almost everything else can be run on a schedule, quietly, in the background, without you touching it.
That’s what we run on, every day, in our own business. And it’s what Claude Code sets up — the schedule, the connections, the safety valves — so you stop paying for someone else’s overhead, and you stop letting a spreadsheet be the thing that stands between you and your own growth.
What To Do First
If you’re building a customizable CMA for your market, the first thing you should ask yourself is not ‘what can AI write,’ but ‘what does a human actually need to touch.’
Almost every CMA starts from the same few sources: a MLS or Realtor.com feed, sometimes a few extra data sets loaded in via API. The mistake is assuming ‘data is the same as the answer.’ It’s not, and it’s getting less true every year.
A row in a table is a fact. A description that’s consistently right across 100 listings is a pattern. A pattern is what you want to hand off to AI, because once it’s built, it works the same way every time, without you re-writing anything when the market changes.
So the first task is never ‘which sentence can AI write.’ It’s ‘what question can I ask the data, that produces a list with a consistent, recognizable shape every single time.’
Almost always, that question is something like ‘give me the last 6 to 12 months of listed homes that closed, in my market, with a sale price and a close date.’ That’s the whole question. It’s not ‘write the description.’ It’s ‘give me a table, sorted by date, with these exact columns.’ And it’s the same question whether you’re building this on your own machine or uploading it to a cloud service.
Here’s what matters more than the specific question: you are the one who defines it, you are the one who checks the first few rows to make sure it’s right, and you are the one who decides whether the output is good enough to hand off to the next step, or whether you need to go back and refine the question before you trust it. Nobody outside your own business gets to touch that definition. You own it, you protect it, because that ownership is what actually makes the rest of the pipeline reliable.
Where to Start
If you want to run this on your own business — not just watch us do it on YouTube — here’s your next step. We offer a free 15-minute call. It’s just us, Al and Victoria, and we’ll look at what you’re already doing and tell you honestly what fits and what doesn’t. The link to book it is right in the description below, and our number’s there too. And if you’d rather start tinkering yourself, the free starter skills are open-source on GitHub — that link’s in the description as well. That’s it. Stop letting a task you dread stand between you and a client. Build the system, and the task goes away.
Frequently Asked Questions
How often does the system update the CMA?
The system runs automatically every time something changes in the market—it doesn’t wait for you to remember or check manually. You get a fresh comp set dated today in your inbox before you even have to think about it, whether it’s Wednesday or Sunday.
Can the AI send emails directly to clients?
No. The system drafts a digest in your voice and schedules it to go out later that day on a timing you control, but you never hand off the actual sending to the AI. That’s a compliance issue we don’t recommend to anyone.
What’s the first step to setting this up for my market?
Start by asking yourself what a human actually needs to touch, not what AI can write. Define the exact data question you’re asking—like “give me the last 6 to 12 months of closed homes in my market with sale price and close date”—then check the first few rows yourself to make sure it’s right before handing it off to automation.
