Can mortgage AI solve the Carol problem?

Sorry for the late-ish email — I closed on a Brooklyn condo Thursday afternoon (5.25% 10-year ARM!), moved all day today & am flying to Chicago Sunday morning. It’s a lot! 

I’ll be sending a detailed Insider Guide to MBA Annual on Saturday, including the full party list 🥳 & an analysis of the attendees & fun things to do in Chicago that a mortgage pro won’t find in the tourist guides (🐢 races!). 

In today’s (free) edition, we’ll tackle VC-backed Valon & Vesta staking their respective claims, PLACE spinning a flywheel & the bureaus fighting FICO w/...irony?

ICE’s new-gen ‘flying V’ competitors are having quite a week, aren’t they?

On the servicing side, a16z-backed Valon raised a boatload of cash at a $2.3B valuation. Meanwhile, Mike Yu & Devon Yang’s Vesta landed a $30M Series B, bringing its total funding to $85M. Vesta, whose clients include Pennymac & NAF, didn’t disclose its valuation, but it's hunting more big-name lenders.

Yu told TechCrunch that demand has “exploded,” w/ revenue up 12x year over year. (Not to brag, but The Mortgage Scoop’s revenue is also up 12x. We started monetizing in mid-Oct. ’25, but still. Call me Ben Horowitz. 😜)

The big selling point for both companies is AI agents that can take rote work off employees' plates & dramatically increase productivity.

At Valon, a customer service employee who might ordinarily handle 700 loans can reportedly manage roughly 3,000 using its AI agent, Ditto. Valon also has a voice AI agent that's freakishly good, right down to the “umms” & intonations. At Vesta, AI agents can handle tasks, track workflows & potentially shave days off the mortgage process. Some lenders are even using Vesta's AI to make underwriting decisions.

Pretty compelling stuff!

But here's the rub: I suspect raising big VC money & getting lenders to adopt the technology might be the easy part. The real challenge is convincing mortgage's legion of conservative, compliance-obsessed lenders to fundamentally rethink how they operate.

Take my own mortgage, which took 48 fucking days to close. (I am not exaggerating.)

For the final week, a woman we'll call Carol was my point person. Getting across the finish line required dozens of frantic phone calls, emergency emails & new conditions appearing at the 5-yard line.

At one point, Carol & the underwriters dug up an address where we lived before COVID & demanded a letter confirming we didn't own the property. There was zero reason to believe we did, but apparently we needed to put that in writing. Because… mortgage?

An AI agent that reviewed the entire file upfront, flagged potential issues & applied a little common sense probably could've shaved a week off the ordeal. Maybe more, I dunno.

But you can give Carol the world's greatest mortgage technology. If her bank still insists on doing things the same old way, what exactly have you solved?

That's the real test for Valon, Vesta & the next gen of mortgage tech: Can they change how lenders work, or just help them do the same dumb shit a bit faster?

Does Ben Kinney have a wealthy benefactor? I’m joking but like, PLACE has been on a shopping spree for the ages (Imelda 🇵🇭 would be jealous). The real estate tech company that owns Envoy Mortgage has picked up Maxwell & Radian's real estate services/title business. On Friday, PLACE announced another acquisition, this time for mortgage recapture platform Ardley (h/t Flávia Furlan Nunes).

So what exactly is Ben & Co. building?

I caught up w/ PLACE President Chris Stuart & newly installed Envoy CEO Chad Smith, formerly of Better, to find out.

The (kinda) short answer: an integrated ecosystem connecting real estate agents, loan officers, title & settlement providers, mortgage technology & institutional clients. The idea is to give top producers better tech, back-office support & access to more business, while bringing down customer acquisition costs.

And PLACE has some awfully big mortgage ambitions. Smith told The Scoop that Envoy, currently running at an annualized pace of nearly $3B in originations, wants to reach $10B within the next few years.

Interestingly, less than half of Envoy's volume comes from PLACE-affiliated real estate teams. Smith wants to recruit top producers who can generate business independently but also benefit from PLACE's relationships.

He said Envoy producers averaged 40% to 44% more loan units over the past 12 months compared to their previous 12 months, gains he attributes to technology, operations, products & access to referral partners.

PLACE is also open to acquisitions & acqui-hires. Smith sees lenders originating $3B or less as a particularly interesting opportunity, though he isn't interested in getting into irrational bidding wars for talent.

"We're not going to just grow to grow to get a LinkedIn post," he said.

There's another intriguing element here: PLACE's institutional business, which includes loan diligence, fulfillment & REO asset management. Stuart wants to connect that deal flow w/ the company's retail real estate & mortgage operations. A different kind of all-in-one.

From the looks of it, PLACE isn't just trying to capture more mortgage referrals from real estate agents, which is what one might expect. It wants to participate in just about every part of the housing transaction, much of it the stuff consumers never see. The messy back-end.

The $100B question is whether all these businesses can actually work together as one ecosystem. Housing has seen oodles of ambitious flywheel strategies, but few (any?) have worked…

FICO wants to cut the credit bureaus out of calculating & distributing its mortgage scores. But the bureaus are like, I’m rubber, you’re glue. Their latest pitch: How about lenders stop buying scores until they actually need them, sucka?!

In the latest escalation of mortgage's increasingly nasty credit-scoring battle royale, TransUnion & Equifax this week rolled out products that let lenders pull a borrower's credit file now & pay for the score later, when there's a better chance the loan will actually close. BNSL! Buy now, score later.

Lenders might dig it. After all, they spend the GDP of a small Micronesian nation pulling credit on borrowers who might 👻 them after the first phone call. The credit file still costs money, & scores will eventually be needed for underwriting & pricing. But like, why pay for FICO upfront on a loan that's going nowhere?

Here's where things get delightfully weird: FICO's Direct License program would let participating resellers calculate & distribute its scores w/o relying on the bureaus to do the scoring.

So… stick w/ me for a minute here. In theory, a lender could buy the underlying credit data from a bureau, skip the score until the loan looks promising & then have a reseller generate the FICO score directly, right? The bureaus keep the data business, FICO keeps the scoring business, the reseller gets its cut & the lender potentially saves some cash. Everyone wins! Maybe!?

I'm admittedly just a simple man who doesn't know whether the technology, contracts, approvals & pricing will actually allow all this to work together. But it's a fun potential resolution to consider: Two sides trying to cut each other out of the transaction might end up helping each other do exactly that. Thus, there is only one conclusion we can reach: Bill Pulte is a genius 🧠 .

(🙏 If you like what you’re reading, tell a fellow mortgage junkie to sign up here.)