Vesta raises $30 million after jump in agentic mortgage automation
Vesta, the AI-native company that helps lenders process residential mortgages, announced on Thursday, October 8, 2026, a $30 million funding round led by Conversion Capital. According to the company, the round comes after usage of AI agents on the platform grew 988 percent over three months in 2026 — and with customers Pennymac and New American Funding participating as investors, alongside Citi Ventures and Andreessen Horowitz.
The company and the round
Vesta was founded in 2020 in San Francisco by Mike Yu (CEO) and Devon Yang (CTO), and offers software that automates the origination of residential mortgages for lenders. Customers include Pennymac, Verus Mortgage Capital and New American Funding, and the platform integrates with systems such as nCino, Blend and Argyle, according to Crypto Briefing.
The $30 million round was led by Conversion Capital, with participation from Citi Ventures and Andreessen Horowitz, reports TechCrunch. Particularly notable is that two of the company's own customers, Pennymac and New American Funding, are participating as investors — a signal that customers see value in the product, in a market where lenders are typically conservative about new technology. Total capital raised now stands at $85 million, according to TechCrunch. Valuation and equity stakes have not been disclosed.
Yu told TechCrunch that the timing was ideal because demand for the product has "exploded over the past year."
The market Vesta is attacking
The backdrop is a market that remains heavily manual. According to the company's own framing, as relayed by TechCrunch, it takes roughly 40 days to close a mortgage in the US, at a cost of around $11,000 per loan. These figures come from the company itself and have not been independently verified.
From assistant to agent
What distinguishes Vesta from earlier loan-workflow automation is the shift from AI as an assistant to AI agents that perform the work — including, at some lenders, credit-policy decisions on loan approval.
Yu describes a deliberately gradual rollout model: "Many of our customers start an AI agent with a person approving its work, then let it handle a share of loans on its own, then expand."
That means humans first review the agent's work, then a portion of loan files are handed over to the agent alone, and autonomy expands as trust is built. Yu says some lenders are already using the agents in credit-policy decisions — but none of those lenders are named in the coverage, and the extent of fully autonomous underwriting decisions among customers is unclear.
Claude Sonnet 4.5 as the technical trigger
The company attributes the recent capability jump to Anthropic's Claude Sonnet 4.5. Yu tells TechCrunch:
"For us, the big breakthrough was [Claude] Sonnet 4.5, which we simply found was much better at following user-configured instructions over the time horizons we need than previous generations."
In a mortgage workflow, where an agent may have to retrieve documentation, assess income, credit and property information, and apply a lender's credit policy across long work sequences, the ability to stick to configured rules over time is a precondition for the agents being usable at all. This is the company's own assessment of model choice, not an independent conclusion.
The numbers — all from the company
Several key figures are circulating around the announcement, and all are supplied by Vesta itself:
- Revenue growth of 12x year over year, according to Yu to TechCrunch.
- Over $100 billion in loans per year that the company has helped originate for lenders, according to Yu.
- 988 percent growth in AI-agent usage on the platform over three months in 2026, according to Crypto Briefing.
- Up to 25 percent lower origination cost at lenders in 2026, as stated by the company.
None of these figures have been independently verified, and they should be read as the company's own account of growth and customer effects.
The competitive picture
Vesta in practice competes on two fronts: against established legacy systems such as ICE Mortgage Technology, and against other AI-native companies such as Xpanse, which is also attempting to automate the mortgage process.
Yu argues that the legacy systems are not built for AI agents — a position that is understandable from a founder, but which points to a real architectural difference: systems designed before the agent era lack built-in support for logging agent reasoning, gradual autonomy, or agent groups working in parallel on loan files.
Accountability and compliance
When AI agents begin making credit-policy decisions in a strictly regulated market, the question of accountability becomes central. Yu's position is clear: the companies remain responsible for underwriting decisions, regardless of which software or which AI agents they use.
He further states that all actions and all reasoning behind a decision are logged, both for compliance and to allow AI decisions to be audited. That means a loan officer or a supervisor should in principle be able to trace why an agent made a given decision.
How these audit and compliance trails work in practice, and how regulators approach autonomous credit policy, is, however, not documented in the coverage. The logging is the company's own description; no independent party has verified it or explained how authorities assess it.
What remains open
The announcement paints a concrete picture of how agentic AI is being rolled out in regulated finance: gradual autonomy, human accountability resting with the lender, and full logging. But several questions remain. Which lenders are actually using the agents for autonomous credit-policy decisions is not known. Whether customers' reported cost reductions hold up under supervision and in loss scenarios is unresolved. And the platform's own performance — the 12x growth, the 988 percent figure, the 25 percent savings — for now rests on the company's own numbers.