For as long as we've sold houses, a brokerage ran on two pillars. Real estate agents and Staff.
Real estate agents brought the deals: the listings, the buyers, the signed contracts. Staff did the paperwork: the transaction coordinators, compliance managers, commission desks, and brokers of record who moved every file from listing to contract to close. Keeping track was tricky, so the industry spent the last century building the back office to manage it. The MLS and the National Association of Realtors arrived in the early 1900s, standardized forms mid-century, e-signature in 2003, then transaction software put the file in the cloud.
Software made the file digital, but the two pillars never changed. Every closed deal still meant a human did the paperwork: opened the file, chased the signature, ordered the title, cut the check. More real estate agents meant more files meant more staff. A broker-owner could glance at the P&L and tell you exactly how it was governed: commission splits for real estate agents, salaries for staff. Every cost was a head you could count.
Then in the space of 24 months, it all changed. The work was no longer constrained by headcount. Intelligence could do the paperwork too. Read every page, chase every signature, draft every follow-up, the moment the file landed. It needed a name, and the one that stuck was Digital labor. And you'd pay for it not by the split or the salary, but by the Close.
Two pillars became three. Real estate agents. Back office staff. And now digital labor: software that does the paperwork, not just stores it. Not a new tool for real estate agents to learn, not another seat on the staff plan, but a third thing that sits alongside both and takes the busywork off their desks. Just one small problem:
It doesn't fit the org chart we spent the last hundred years drawing.
The sampled-file blind spot
Boil down a century of brokerage operations and it's really three questions:
Is the file complete?
Is it compliant?
Will it close on time?
Imagine a brokerage with 40 real estate agents. The compliance manager has all three answers by lunch. Open. Check. Clear. The old files reviewed themselves.
Now picture the same manager. Her brokerage recruited well, growing to 400 real estate agents, 3,000 closings a year, 100-plus documents in every file. Same three questions. What does she find? A queue. A sampling rate. One in ten files actually read. She has no idea.
And the real estate agents? They did everything right. The file is in DocuSign. The addendum is in a Gmail thread. The inspection report is in Dropbox. Nobody dropped the ball; the ball is just in five places, and someone back at the office has to find it.
What's going wrong? Well, files are not invoices.
They don't live in one system; they cross all of them: DocuSign, Gmail, Outlook, Drive, a text thread. There's no monthly cadence you can plan around; deals move daily. There's no stable document you can sign off on; a single addendum can change the whole file overnight.
It's like a home inspector who checks one room, signs the report, and hopes the other eleven look the same.
You've seen the headlines. A $418 million settlement over commissions. A $1.8 billion verdict before it. Disclosure claims that surface two years after the keys changed hands.
That's the loud real estate story. If you're a broker reading this, I imagine your instinct is to add staff, or add checklists for your real estate agents to fill out.
But before you do, let's zoom out.
The quiet AI story: cheaper and smarter
It is better to be directionally right than precisely wrong.
AI will keep getting cheaper. In 2023, matching GPT-4-level intelligence cost $60 per million tokens. Today, about 40 cents. A full transaction file now costs less to read than the stamp on the disclosure packet.
AI will keep getting smarter. In 2023, GPT-4 solved 2 out of every 100 real software bugs. Today's leading models: 94. A model that can fix code can certainly notice a missing initial on page 31.
Cheaper and smarter, every month
Cost of frontier-level intelligence, USD per million output tokens
Share of real software bugs AI can solve
This is the shape of technology. Title searches, lockboxes, e-signature. You don't know the price, or exactly when, but you can see the direction.
Here's what changes when every file is read. The reviewer who used to sample now starts her day with the files that actually need her. Clean files clear on arrival, missing signatures get chased the same day, and the exceptions that used to surface the week of closing show up while there's still time to fix them.
That's what we're building with every brokerage we work with: review that covers every file instead of a sample, exceptions caught on day one instead of day thirty, and growth that doesn't come with a hiring plan attached.
The problem isn't paperwork. It's who's still doing it.
So if the file is usually fine, what's the issue? Well, in one word: people. Not that you have the wrong ones. That the paperwork still runs through them at all, and the first thing that breaks when it does is review. Call it sampling.
You don't know which file has the problem.
One hundred files. Two ways to review them.
Illustrative. Assumes 80% of files are clean, as in the example below.
Ten files opened. Ninety never read.
Eighty clear on their own. Twenty reach a reviewer.
A lender has underwriting. An insurer has actuaries. These are disciplines you can measure risk against. All a sample can tell you is that the ten files you opened were fine, not whether the ninety you didn't were.
Finance says, "we can't add another coordinator", the broker of record says, "we can't not" and you don't know who's right because there is no shared language of risk.
For example, right now, every file, no matter how clean, waits for the same human. Meanwhile the real estate agent waits for the human, and the client waits for the agent. But a listing agreement with 18 of 18 documents signed doesn't need the broker of record. A purchase file missing its lead-based paint disclosure does.
In isolation this seems trivial, but think again. If 80% of your files are clean and you can route them past the queue, the reviewer's judgment goes five times further on the 20% that aren't. You just quintupled compliance capacity without a job posting. Not every file needs the broker.
And review is just the first job. The same logic runs through the whole transaction: the missing-document chase, the title order, the inspection scheduling, the commission disbursement. All of it is paperwork. None of it needs a person. The judgment calls and the client relationships do.
This is the future of brokerage operations: the paperwork goes to software, the judgment stays with people. Not rubber-stamping files or sampling them, but allocating human attention with the same rigor brokerages apply to splits, recruiting, and marketing spend.
Fortunately, that's what we're building.
"The AI back office" (since 2026)
But how do you hand paperwork to software? There are thousands of files. They arrive in five systems. They cut across every office and every stage of the deal.
At Joymore, our digital labor is called Joy, and we like to say, "start it, understand it, do it".
Start it: Paperwork is cheapest to get right at the beginning, so that's where Joy starts. A real estate agent texts Joy the terms from the kitchen table: address, price, close date, contingencies. Joy drafts the contract on the right forms, fills it from what she already knows about the listing and the parties, and sends it for e-signature. The file is born complete. Prefer to keep your current process? Joy reads every file instead, pulling each contract, disclosure, and addendum from DocuSign, Gmail, Outlook, Drive, and Dropbox the moment it lands. Either way, real estate agents keep working exactly where they already work. Nothing new to log into or check off.
Understand it: A document is just a PDF until you know what it means for this deal. Joy knows what the state requires, what the county and city add on top, and what your brokerage insists on beyond that: required documents, signature rules, disclosure timing, closing standards. That stack of requirements used to live in a veteran coordinator's head. Now it's defined once and applied to every file, in every office, the same way.
Do it: Judgment is baked in. Clean file? It clears on arrival. Missing a signature? Joy drafts the follow-up in the real estate agent's inbox, so the agent hears about it the same day, not the week of closing. Contract signed? Joy opens the next step before anyone asks. A real exception? It routes to a human with the context to decide. And you pay only when the deal closes.
Same paperwork. Two bills. Your choice.
A page of paperwork done by AI keeps getting cheaper. The same page done by a person keeps getting more expensive: more documents per file, higher wages, more rules. The space between them is the routing decision every broker is making, whether they know it or not.
The gap is what we're building Joymore to close.
None of your coordinators went to school for AI operations, and none of your real estate agents signed up to learn another system. So, we're making it as easy as possible: Joy works on the files real estate agents already create, in the tools they already use. Zero adoption required. The best real estate agents never notice Joy is there. They just notice their files stopped getting stuck.
Where's this heading? We start at the beginning of the file because paperwork that's right on day one never needs rescuing on day thirty. But the file is only the first stretch of the road. A transaction is a chain of hand-offs, listing to contract, contract to close, close to commission, and every link in it is paperwork. Today Joy writes the contract, clears the file, and keeps it moving. Next she chases the signature, orders the title, schedules the inspection, and prepares the disbursement, just like your best coordinator. Then everything around the deal that requires intelligence. Except she will do it on every file, on the day it arrives, for every real estate agent in the brokerage at once.
The end state is simple to say: AI does all of the paperwork. People do the parts that were never paperwork to begin with.
So, how do you 'manage' intelligence? Files are just data. Digital labor is just a hire. A back-office hire, not a replacement for the real estate agent whose name is on the sign. And we're building a back office ready to run on both of them.
Don't miss the decade
In May 2025, Nvidia CEO Jensen Huang put it plainly: "You're not going to lose your job to an AI, but you're going to lose your job to someone who uses AI."
Swap "job" for "agents" and you have the real estate version. Eighteen months on, the only debate in real estate is how to staff for it. If you spend too much time online, the choice appears to be:
"Recruit hard? Blow up compliance. Stay safe? Lose your real estate agents to the brokerage that didn't."
And I'm here to tell you that's a false choice. The brokerages pulling away aren't hiring more staff, or fewer, they're taking the paperwork off people's desks. Every file read. Every next step started. And their real estate agents feel it: faster clears, fewer "where's the addendum" emails, more time in front of clients.
Your paperwork is the biggest cost in the brokerage that nobody owns. It's also your single greatest opportunity.
The third pillar is here. The broker's job just got bigger. And yes, we're here to help.
Back to work.


