Why would a brokerage build an AI system instead of letting agents use AI on their own?

Because an agent using AI on their own builds their own thing and takes it with them. A brokerage that builds shared skills owns the asset. Every agent runs the same CMA, the same listing presentation, the same follow-up, and the quality floor across the whole office comes up at once. It stops being 60 agents doing 60 different versions of the job and becomes your brokerage's way of doing it.

How does an AI system help with recruiting agents?

It gives you something to show that isn't a split. Most recruiting conversations are desk fees and commission structure, which every brokerage down the street is also offering. Opening a system and showing an agent that on day one they can produce a CMA, a listing presentation and a single-property website that look better than what they build today is a different conversation, especially with a producer who already knows AI is coming and doesn't want to build it alone.

Does an AI operating system actually improve agent retention?

It's the strongest retention mechanism most brokerages have available. An agent whose daily production runs on skills your brokerage built has to rebuild a working business to leave. That's different from a split, which a competitor can simply beat. The agents most worth keeping are the ones who use it most, so the lock strengthens exactly where you want it to.

What does a consistent client experience have to do with AI?

Right now every agent in your office runs a CMA differently, presents differently, and onboards a client differently. The brokerage name goes on all of it, but there's no shared standard behind it. Build the CMA once, correctly, and now every seller who meets any of your agents gets the same document, the same depth, the same look. That's what turns a roster of individual brands into a brokerage brand.

What should a brokerage build first?

The document your agents produce most and do worst. For most offices that's the CMA, because it decides listing appointments and it's where the quality gap between your best and weakest agent is widest. Get that one right, prove the outcome, then add the listing presentation, the single-property website and the follow-up sequences.

What if only a handful of my agents actually use it?

Start with them. Nothing requires you to roll it out to the whole roster on day one, and the build cost is the same whether ten agents run it or sixty, which is an argument for building fewer skills well rather than for seating everybody. Measure adoption from the first month so you know whether to expand instead of guessing six months in.

Claude Cowork for Real Estate Brokerages: Recruit and Keep Better Agents

Most brokers I talk to have thought hard about using AI to recruit and haven’t thought at all about using it to run the brokerage.

That’s the wrong way round, and it’s costing them the recruiting outcome they actually want. The thing that makes a recruiting pitch land is having something real to show, and the only way to have something real to show is to have built it for your own operation first.

I’m William Zhang, founder of Real Estate AI Society. I help real estate brokerages transition to become AI-first and AI-native in their market. The conversation is remarkably consistent, so this is what brokers are actually trying to fix, and what changes when they fix it.

The Pain, Named Honestly

Your recruiting pitch sounds like everyone else’s. Splits and desk fees. A producer weighing three offices is comparing three versions of the same sentence. Nothing in that conversation is about what they can do on Monday that they can’t do today.

Your best agents have nothing holding them. A better split is easy for a competitor to offer. If the only thing keeping a top producer is economics, you’re one recruiting call away from losing them, and the ones you lose are the ones you can least afford to.

Every agent does the job differently. One agent’s CMA is a polished document. Another’s is three Zillow screenshots. Your name is on both. A seller who interviews two of your agents gets two different companies.

You are the ceiling. The broker is usually the best person in the building at the listing appointment, the pricing conversation, the negotiation email. That skill lives in one head and doesn’t scale, so the brokerage grows only as fast as you can personally coach.

The work you do yourself is slow. Brokers routinely spend three or four hours building a CMA they’re proud of, and then build the next one from scratch. Same for listing presentations, lease packages, inspection responses.

Training doesn’t stick. One big AI workshop and nothing changes by Thursday. Everybody’s seen this happen.

And clients have started asking AI who to hire. Buyers and sellers now open an assistant and ask who they should work with in their market. Whoever published something specific gets named. Most brokerages have no answer to this at all.

What Changes: Four Outcomes

1. Recruiting Conversations Stop Being About Splits

When you can open a system and show a prospective agent the CMA, the listing presentation and the single-property website they’ll be able to produce on day one, you’re no longer competing on economics. You’re offering capability.

The agents this lands hardest with are the good ones. A producer doing real volume already knows AI matters and already knows they don’t have time to build it themselves. “We built it, and you get it the day you join” answers a question they’ve been quietly worried about.

That’s a differentiated pitch in a business where almost nobody has one.

2. Your Top Agents Have a Reason to Stay

This is the outcome brokers underrate most.

An agent who runs their CMAs, their listing presentations, their social content and their follow-up through skills your brokerage built has their working life inside your system. Leaving doesn’t mean changing a sign. It means rebuilding a business that currently works.

Compare that to a split, which is a number a competitor can simply beat.

And it compounds in the right direction: the agents who use the system most are your producers, so the hold is strongest exactly on the people you most want to keep. Retention stops being a conversation you have after someone’s already interviewing.

3. One Consistent Client Experience, Which Is Your Brand

Today your brokerage’s brand is whatever each individual agent happens to do. Sixty agents, sixty standards.

Build the CMA once, properly, with your pricing logic and your branding, and every seller who meets any of your agents gets that document. Same for the listing presentation. Same for the onboarding sequence a new client walks through.

Your newest agent presents at the level you set. Your weakest agent’s listing appointment stops being a liability with your name on it. A seller who interviews two of your agents sees one company.

That is what a brokerage brand actually is, and most brokerages don’t have one because they’ve never had a mechanism to enforce a standard. This is that mechanism.

4. Your Agents Win More Listings, Which Is Your Revenue

The house gets paid when agents produce, so the honest measure of any of this is whether it wins more business.

A few things move that needle directly:

CMAs clients understand. The reason to build this well isn’t speed, it’s that a clear, well-structured CMA can carry the pricing conversation on its own. Sellers can reach a decision from the document instead of needing to be talked into a number in person. That shortens the path from appointment to signed listing.

Listing presentations that don’t take a day. Built from one prompt, with real market data, in your brokerage’s format.

A single-property website for every listing. It used to be a luxury reserved for expensive homes. As a brokerage skill it’s standard on every listing, which is a real thing your agents can promise in a listing appointment that the agent across the table can’t.

Social content that actually ships. Most agents know they should post and don’t. A content skill plus a scheduling tool means it goes out whether or not they felt like writing it that week.

Follow-up that matches the lead source. A referral and a cold internet lead need different sequences. Most agents run one generic drip, or none. Built once at the brokerage level, every agent gets sequences per funnel.

And getting named when a client asks an AI. This one compounds for the brokerage and the agent at the same time, and almost nobody is doing it yet. I wrote up how an AI recommendation produced an actual client for me, and the mechanism is repeatable at an office level.

Why This Is a Brokerage Build, Not an Agent Purchase

Every one of those outcomes depends on the work being shared.

If each agent buys their own AI subscription, you get sixty private experiments of wildly varying quality, no standard, no brand consistency, and nothing that stays when someone leaves. The good ones build something and take it with them.

Built at the brokerage level, the skills are the brokerage’s asset. One person builds the CMA correctly and everyone runs it. That’s the entire difference between AI as a tool your agents happen to use and an operating system your brokerage owns.

If you want the nuts and bolts of what gets built, an AI operating system for a real estate brokerage breaks down the five skills worth starting with. If the training half is what you’re weighing, why one big workshop never sticks covers the cadence. And to see the shape of it at the individual level first, the Cowork beginner build walks through the setup end to end.

Where to Start

Pick the document your agents produce most and do worst. For most offices that’s the CMA. It decides listing appointments, and it’s where the gap between your best and weakest agent is widest.

Build it properly, once. Your comp logic, your format, your branding.

Train in a series, not one session. Monthly beats a single workshop badly. People need to try it, fail at it, and come back with a question.

Measure adoption from month one. Who’s using it, what they’re producing. Without that you’re guessing about the exact thing you were worried about at the start.

Then add the next skill. Listing presentation, single-property website, follow-up sequences, negotiation analysis.

The brokerages that win 2027 recruiting aren’t going to win because they picked the right AI model. They’ll win because they spent this year building the documents their agents produce every week, while everyone else was still deciding.

Talk to Me About Your Brokerage

If you’re weighing this for your office, tell me what you’re trying to fix and I’ll come back with something specific.

Start here. It takes about two minutes, and it asks whether you’re after training for your agents, the system build itself, or both, so the conversation starts in the right place.