Vesta, the San Francisco company that sells mortgage lenders a loan origination system built from scratch, said on Oct. 8 that it raised a $30 million Series B round led by Conversion Capital. The notable part is who else wrote checks: Pennymac, New American Funding, Citi Ventures, nbkc bank, First American and FirstKey Mortgage — lenders and settlement firms that are also Vesta’s customers or partners — alongside Andreessen Horowitz, Zigg and Navitas. The round brings total funding to $85 million.
In its announcement, Vesta said revenue has grown more than 12 times over the past twelve months, and that customer count and headcount are each up more than 50%. Lenders using or converting to the platform originate more than $100 billion in loans a year, the company said, while adding that fewer than 5% of the market runs on Vesta today.
What the agents actually do
The claim at the center of the round is operational rather than aspirational. Vesta said its AI agents have been in production all year — “scrubbing applications, making underwriting decisions, and reviewing closing documents” — and that about 40% of tasks completed on its platform are now handled by AI agents and automated workflows, a share it says is climbing.
Vesta argues mortgage is unusually well suited to agentic automation for three reasons it lays out in the post: the work of originating a loan is repetitive but hard to structure with conventional software; the industry already has verification and quality-control loops, because the largest loan buyers require a share of closed loans to be audited, giving a built-in accuracy metric; and lenders track manufacturing cost per loan closely enough that replacing manual effort produces a return that can be measured rather than asserted.
The company also makes a technical argument about why it had to rebuild the underlying system of record before adding agents. Vesta said the platform most of the industry still runs on is more than twenty years old, uses different backend fields for the same value on different screens, and allows only one party — human, machine or integration — to edit a loan file at a time. Building agents on top of that, the company said, would mean waiting for a human to leave the file and driving a Windows application through its interface. Vesta’s co-founders, Mike Yu and Devon Yang, both worked at Blend before starting the company, and spent roughly two years writing the core platform before converting lenders onto it.
Customers as investors
The investor list is the part worth watching. Pennymac selected Vesta in September 2025 and, according to Vesta, has reported 25% back-office efficiency gains with a plan to automate 80% of its production workflows by the end of 2027. That figure comes from Vesta’s own announcement and has not been separately confirmed by Pennymac, whose second-quarter profit fell 84% as higher rates squeezed origination volume. New American Funding, which announced a partnership with Vesta in July 2026, has separately been deploying voice AI to field customer calls.
That overlap cuts both ways. Strategic money from the companies that buy the product is a strong signal of adoption, and it also means several of Vesta’s reference customers now have an interest in the vendor’s valuation. On our reading, the better independent test will be whether the back-office savings show up in lenders’ own reported cost-per-loan figures over the next several quarters.
Where the bottleneck sits
Vesta was candid about the constraint on growth, and it is not demand. Standing up a system of record today, the company wrote, “means translating how a lender works into thousands of settings, one at a time, by hand, and it takes the better part of a year.” Its stated plan is to have an agent learn a lender’s procedures by reading them, asking questions and running through past loans, so that configuration stops being the slowest step in a conversion.
The company’s long-range pitch is that full automation would compress closing timelines to the seven-day minimum required by law rather than the 45 days it says is typical now, and make categories of lending that are marginal at current costs — small-balance loans, manufactured-home loans, incremental refinances — economic again. Vesta said its customers will originate the first trillion dollars of agent-manufactured mortgages within five years. That is a projection, not a result.
The round lands in a stretch when mortgage technology capital has been concentrating on automation rather than consumer-facing tools. Sela raised $21 million for AI agents that work mortgage sales calls, and proptech rounds across adjacent sectors have followed the same logic: sell into a known cost line, not a new one. Vesta raised a $30 million Series A led by Andreessen Horowitz in January 2022; it has taken nearly five years for the follow-on.
