Top 3 Claude Skills for Realtors (The Ones That Survived 2 Months)

I gave Claude three jobs in my real estate business and ran them for two months on real clients and real money. Most of what I tried got deleted. Three survived — and one of them drafted the follow-up that helped me close a $500,000 deal. This post is about the Claude skills for realtors that actually stuck, and the rule I used to decide which ones lived and which ones died.

Quick background. I’m William Zhang, a full-time real estate agent in Austin, Texas, and the founder of Real Estate AI Society. I did about $6.5 million in production volume in 2025, and roughly 95% of it came from YouTube. I started building my AI assistant in Claude on March 13, and I’ve been refining it ever since. Yesterday I shared the honest 2-month review of all the workflows I tried. Today I want to zoom in on the three skills that earned a permanent spot, and how they actually fit together as one system instead of three loose tools.

Here’s the rule I used to decide. A skill had to add real business value, get tested on actual clients with real money, and save me time on something I’d otherwise do by hand. If I used a skill a few times and it didn’t pay back the build, I cut it. Some of what I built ended up in that pile. These three didn’t.

How My AI Real Estate Operating System Fits Together

Before the three skills, the part that matters most is the shape. I think of this as an operating system, not a folder of separate tools. Claude is the foundation layer. On top of that sits one operator — my executive assistant skill, which I named Ashley — and Ashley orchestrates the two specialist skills underneath her: the CRM skill and the YouTube production skill.

Why this matters: I don’t open three different tools in the morning. I open one. I type “good morning” into Claude Code, and Ashley pulls in everything else — the CRM activity, the YouTube stats, the leads I should reach today, the one thing I should focus on. That’s the difference between an AI tool and an AI operating system. The tool answers when you ask. The system runs while you’re not looking.

I broke down the deeper setup of the executive assistant pattern in my AI executive assistant for real estate. This post is what got built on top of it.

Skill #3: The YouTube Production Skill

I’m a content-first agent. About 95% of my $6.5 million in production volume in 2025 came from YouTube and social media. Which means the bottleneck on my business isn’t the camera — it’s everything that happens around the camera. Thumbnails, titles, hooks, SEO descriptions, slides, uploads. That stack used to eat one to two hours per video on the post-production side, and I was doing two or three videos a week. The math didn’t work.

The YouTube production skill packages all of it. It generates the title and thumbnail concept using best practices I’ve coded into Ashley, drafts the hook and the SEO description, builds the slide deck for the video (the slides you see in the video on this page were made by Ashley), and handles the upload itself. The uploading part is a separate skill underneath, but Ashley calls it for me. I just say “upload this one.”

Here’s the part I didn’t expect. Once the prep stack was off my plate, I started spending that hour on the creative side instead — the story, the angle, the concept. The video got better because the packaging stopped competing with the thinking. That’s the real return on this skill, more than the time saved. I’m not racing post-production anymore, so I’m not cutting corners on the content.

Skill #2: The CRM Follow-Up Skill (Follow Up Boss)

This is the money skill. One of the texts Ashley drafted through this workflow turned into a $500,000 deal that I’d otherwise have let slip. That alone paid for the entire system.

I use Follow Up Boss as my CRM. It’s powerful, but powerful in the way that means there are a lot of features you have to learn to set up correctly, and if you don’t set it up right it just sort of sits there. I never had the patience to fully master it. So I gave Ashley the documentation and told her to manage it for me.

Now when I run the “good morning” command, the CRM skill pulls every lead and every recent activity, scores each lead hot, warm, or nurture, and drafts the exact text or call message I should send today — in that lead’s language, with the right context from their last few notes. Then it creates the daily plan: call this person, text that person, follow up on this deal. I copy and send, or click and call.

The next thing I’m building on top of this is an AI voice calling agent. When a new lead comes in, the voice agent calls them as an appointment setter, and if they’re not ready, Ashley keeps the nurture sequence running on her own. I share the build progress with Real Estate AI Society members — if you want the prompts for the version of this skill I run today, the free starter kit through the newsletter has them.

Skill #1: The AI Executive Assistant for Real Estate (Ashley)

This is the one I almost cut. When I first built her, Ashley looked like a glorified to-do list. Cute, marginally useful, not worth the build. I came close to deleting her in the first two weeks.

What changed is that the more I used her, the more she learned how I actually run my business — what I think is important, what I skip, when I’m in client mode versus content mode, and what “today’s one thing” should look like in each. She stopped feeling like a to-do list and started feeling like a partner who already knows the context. She doesn’t ask “what do you want to do today” — she tells me, and asks me to confirm.

That shift is the whole reason this is now my #1 skill. Without Ashley as the operator, the CRM skill and the YouTube skill are still just tools. With her, they’re a business that runs in the background. The executive assistant pattern is the most important piece of the whole stack, and it’s the one most agents skip because it’s the least obvious one to build first.

If you want to set up your own version, the prompt I use is in the free starter kit through the newsletter, and I cover the architecture in the Real Estate AI Society curriculum.

What Claude Still Can’t Do for an Agent (The Honest Part)

There’s an honest limit to what Claude AI for real estate agents can do today. This system doesn’t close deals. It doesn’t hop on a call with a seller. It doesn’t walk a buyer through the foundation issues in a South Austin teardown. It doesn’t do the part of real estate that actually generates revenue — that’s still me.

What it does is take the back-office work that doesn’t generate revenue but still has to get done. Follow-ups I used to skip. Thumbnails that used to take an hour. Daily plans I used to sketch on a notebook and lose by 2 p.m. That work doesn’t disappear when you don’t have an assistant — it just delays the part of the business that actually pays.

I’m not pretending this replaces the human side of the job. I’m saying it gives you back the four or five hours a day that the human side never got because the admin ate them first.

Why This Compounds

Here’s the part I keep telling other agents. The productivity gain you see on day one is real but small. The real gain shows up a few weeks in, after you’ve started stacking skills on top of each other and Ashley has learned how you actually work. That’s when this stops feeling like a tool and starts feeling like infrastructure.

The agents, brokers, and team leaders who get fluent in this now are going to compound year over year. The ones who wait are going to keep doing the admin themselves while the AI-native agents close more deals on the same hours. That’s the bet I’m making with my own business.

If you want the prompts for all three of these skills — the YouTube production skill, the CRM skill, and the executive assistant — they’re in the free starter kit. Subscribe to the newsletter and you’ll get them, plus the full curriculum if you want to build the whole system from scratch. The deeper how-I-built-it walkthrough is in Claude for real estate, and the full list of tools I use sits at /tools/.