The Other Forty Days

Add up the fastest demonstrated time for each of those steps and you get less than two days. Federal disclosure rules add a floor of about a week: closing can’t happen sooner than seven business days after the first loan disclosures go out. Yet the average financed purchase still takes five to six weeks from application to close, and it has for as long as anyone has been measuring. The fastest monthly average ICE has ever recorded, 36.8 days in March 2026, is still more than five weeks. A decade ago the same figure ran seven weeks (Ellie Mae, 2016).

They go into the seams. A home sale runs through five separately licensed industries: about 1.4 million agents across more than 300,000 real estate firms, roughly 4,800 mortgage lenders, more than 6,000 title and settlement companies, tens of thousands of appraisers, and hundreds of MLSs. Not one of them can close a home sale without the other four. Nobody is positioned to see the whole transaction. So the transaction is run by its handoffs: a forwarded PDF, a voicemail, a “just checking in” email, a deadline living in someone’s memory.

And that’s only the fragmentation you can see from the outside. Inside the lender, the file passes through a loan officer, a processor, an underwriter, a closer, a funder, and a post-closing team. Inside the title company: a searcher (often a third-party abstractor the title company itself hires), an examiner, a processor, an escrow officer, and a post-closer, with a closing attorney over the top of them in roughly twenty states. Every one of those handoffs is a person keeping a list current by hand, and nobody outside the building can see which desk the file is sitting on, whose queue it’s stuck in, or who hasn’t started yet. When a deal stalls, everyone finds out the same way: somebody finally asks.

The industry’s true API is the email inbox. And it shows:

We pay for the seams in time, in fraud, and in the quiet daily anxiety of every buyer who has ever asked their agent, “Any updates?” and gotten silence back.

The industry has produced two big answers, and both prove the problem more than they solve it.

Answer one: own every seat at the table. Rocket spent roughly $16 billion in 2025 buying Redfin and Mr. Cooper to assemble search, brokerage, lending, title, and servicing under one roof. Zillow is building the same thing from the portal side. Here’s what the walled garden actually proves. Even after buying the brokerage outright, Rocket attaches its own loan to fewer than half of Redfin’s financed buyers, a roughly 45% attach rate as of its most recent earnings call. By our arithmetic on Rocket’s own disclosed volumes, deals that run its agent, its loan, and its title end-to-end amount to at most a single-digit share of its closings. Vertical integration hasn’t solved coordination inside one company, and five of six American mortgages will never live inside those walls anyway. Consumers want choice, agents want independence, and lawmakers (five US senators so far) have started asking hard questions about deals like these.

Answer two: build one platform and make everybody log in. This is the proptech graveyard. First American put over $200 million into Endpoint; it was folded back into the mothership. Doma went private. Voxtur filed for creditor protection. Every “one platform for the whole transaction” failed the same way: it secretly served one seat at the table, so the other seats never showed up.

Meanwhile, today’s best-funded AI companies in real estate are all doing something narrower: automating their own seat. An AI escrow officer. An AI title roll-up. Agentic software inside the title company’s four walls. All real, all useful, and all confined to one seat, because each one’s business model requires owning it.

Nobody has built the layer above the seats.

Ratifyly starts from a different premise: the parties aren’t broken. The seams are.

So we don’t ask anyone to change. The agent forwards the contract the way they’d send it to a transaction coordinator. The lender keeps working in its loan system, and where a lender connects that system to ours, milestone updates flow in with no email at all. The title company keeps emailing the way it always has. Ratifyly reads everything they send, every page, including the scanned addendum with handwriting in the margin. It keeps one accurate picture of the deal and watches every deadline in the contract, the loan, and the closing. Everyone sees the same live transaction: the agent, the broker, the client.

And a licensed person signs off on every call that carries consequence. That is not a disclaimer. It is the design. We’re asking people to trust software inside the exact inbox where the spoofed-title wire fraud lives. That’s why a human keeps the judgment.

This isn’t a plan to write anyone out of the deal. On the transactions that still need an appraisal (most of them), orchestration means the appraiser gets the assignment on day one instead of day twelve. The title processor gets a complete, pre-read intake package instead of a half-forwarded email chain. The transaction coordinator stops chasing paperwork and starts closing more files.

This is the pattern that has already worked everywhere else fragmentation was the disease. Plaid didn’t replace the banks; it connected them, stayed neutral, and became infrastructure. In our own industry, Snapdocs aggregated the notaries nobody wanted to serve, stayed neutral between lenders and title, and now touches one in four US mortgage closings. But it stops at the closing table. The other forty days of the transaction are still coordinated by forwarded PDFs and phone tag.

Those forty days are what we’re taking.

Three years ago, no machine could reliably read a scanned, handwritten closing package, which is why every earlier tool stopped at storing documents a human still had to open. That changed in the last twenty-four months. The newest AI can finally read a bad scan of a ratified contract, handwriting in the margins and all, reliably enough to act on, as long as a person signs off.

At the same moment: the GSEs rebuilt the appraisal as structured data (mandatory this November) and unbundled on-site property data collection to trained, background-checked collectors, which means the valuation can finally run on the same clock as everything else. eNotes hit a record 15% of MERS registrations in January, passing three million cumulative this spring. RON went legal almost everywhere. Every rail this industry needs already exists.

The day a contract is ratified, everything that can start should start: the title order, the loan file, the valuation, the inspection, the HOA resale packet, the scheduling. In parallel, not relay. Every deadline orchestrated, not just watched: software that never sleeps sends the reminder, chases the missing document, and keeps each task moving on behalf of the people in the deal. Every party seeing the same live transaction. Every document read the moment it lands.

The title side shows how close this future already is. Several national vendors will now return a title search the same day, almost anywhere in the country. And when I ask closing offices whether they would accept a search ordered by someone else, most say yes. Nobody does it, because nothing exists to order that search the moment a contract is ratified. That is an orchestration gap, not a technology gap.

The destination is a two-week close available not just to the customers of one company’s walled garden, but to anyone, through the agent, lender, and closing office they already chose.

We’re not building the platform everyone has to use. We’re building the layer that finally makes the ones they already use work together.

Real estate doesn’t need fewer participants. It needs one connected transaction.