Hey guys, just added the deal desk workshop recording to the workshop classroom. Feel free to check it out. If you guys have any questions, don't hesitate to reach out. I am here to help you guys implement AI into your business so you can get back your time.

Hi Guys! Just a reminder that workshop is this Friday we're gonna create a AI Agent that analyzes deals. Pull County Records and do a full analysis don't miss out lock it into your calendars

You are my deal desk. I'm going to give you a property, and you're going to help me decide: pass, dig deeper, or strong opportunity. Work in this order and label every section. 1. WHAT I GAVE YOU Restate only the facts I provided. Nothing else. If you looked something up, name the source next to it. 2. WHAT YOU NEED List every input that would change the verdict and that you don't have. Ask me for them one at a time, most important first. Do not guess any of them. Do not move to step 3 until I've answered. 3. THE MATH Show your work. Formulas visible. Run it as a range, not a single number — low, expected, and high on both ARV and repairs — and show me how the verdict changes across that range. 4. THE TAX LINE Flag that the seller's exemptions (homeowner, senior, senior freeze) disappear when the property transfers. The tax number on record today is not the number the next owner pays. Tell me what that swing does to the deal. 5. VERDICT PASS / DIG DEEPER / STRONG OPPORTUNITY — and the one reason why. 6. THE 5 THINGS TO VERIFY What I need to confirm before I make an offer. Rules you follow every time: - Ask my exit strategy before you run any math: wholesale, fix and flip, buy and hold, or creative terms. The math is different for each one. Never default to one. - Never present an automated estimate (Zestimate, AVM, any algorithm) as a comp. Say what it is. - Comps: last 3 months, quarter-mile radius. If that returns fewer than 3 usable sales, widen the TIME first — go to 6 months. Never widen the geography first. Never cross a municipal boundary, a highway, or a rail line. Tell me exactly how many comps you found and exactly what you relaxed to get them. - If a number is mine to go get — interior condition, repair cost, anything that needs eyes on the property — say so and stop. Don't fill it in. - Label anything you estimated as an estimate. Never present one as a fact. Here's the property: [address, beds, baths, square feet, lot, year built, asking price, anything else you know — even if it's not much]

Friday, August 14 · 12:00–2:00 PM CT · Live on Zoom Every institutional buyer has a deal desk. A team that runs the numbers, pulls the comps, and says go or no-go before anyone wastes a week on it. In two hours we're building yours. WHAT WE'RE BUILDING An agent that takes one input — an address — and comes back with the county record, the assessed value, the real tax picture, and the comps. Automatically. From public data, so every number is verifiable. I ran it on a live off-market deal in Calumet Park last week. It pulled the full county record, caught a tax exemption issue that would've cost the buyer three to four grand a year after closing, and surfaced two renovated sales on the same street that every automated estimate had missed. That's what you're walking out with. Not a demo. Yours, running on your machine, pointed at your market. WHAT YOU WALK AWAY WITH - A working Deal Desk agent on your own computer - Live public-record lookups — assessment, tax rate, property characteristics - A comp engine with real rules: how tight to draw the radius, how far back to go, and which boundaries never to cross - The tax traps that kill deals after closing, and how to catch them before - Your own buy-box math, running automatically on every address you feed it WHO IT'S FOR Agents — price a listing in minutes with defensible numbers instead of a gut call. Investors— underwrite before you drive out there, and stop wasting Saturdays on dead deals. Same desk. Different question. HOW TO PREPARE — READ THIS PART You are coming to build. Not to install. Everyone who shows up with a blank machine slows the room down, and the person it hurts most is you. So before Friday: 1. Go to the classroom and open How to Build Your Agent. Work through it. 2. Install Claude Code. It's in the course, one line, about five minutes. 3. Build your first agent — the one that audits your email and hands you back the deals you let go cold. That's the whole course, and it's about 45 minutes. 4. Show up Friday with it running.

Two hours, live. My real terminal, my real inbox. No slides pretending to be a demo. If you missed it or want it again, the replay is in the classroom. What we covered: An AI agent is a folder. Not something you download. Brain file, skills, sub-agents, hooks, autonomy. Five pieces, and you assemble it yourself. It's not what AI tells you to do. It's what you tell AI to do. Define your process first. AI is the rocket fuel — but you don't put rocket fuel in a broken rocket. The five things everyone gets wrong about agents, including the big one: the model isn't the magic. The context is. Claude, explained simply — chat vs. code, Haiku vs. Sonnet vs. Opus, and why projects keep your context clean. Three live demos: - My assistant read two weeks of calls and told me where I dropped the ball on a live lead - A deal analyzer pulled two off-market deals out of my inbox — the real acquisition basis, not the sticker price — and flagged what it didn't know - Claude took over my browser and replied to every comment on a Facebook post while we watched Then everybody built one. Five questions, and you walk out with your business brain and your first agent named — the one built around the task you hate most. The whole stack runs for under $100 a month. Notion free. Supabase free. GitHub free. Fathom free. Claude's the only real line item. The brain-file prompt is in the classroom too. If you didn't finish it live, go run it — take your time, and do it more than once. New workshop every month, free in here. You are the creator of your reality. — Coach Ma