Vesta Raises $30M to Deploy AI Agent Swarms in Mortgage Lending

AI Agent Swarms are moving from research demos into the regulated core of American finance. Vesta, a San Francisco startup building an AI-native loan origination system, announced on Thursday that it has raised a $30 million Series B round led by Conversion Capital, with the money earmarked to scale AI Agent Swarms that handle document review, data validation, and even underwriting decisions for mortgage lenders.

The round brings Vesta's total funding to $85 million, according to reporting by TechCrunch. Conversion Capital, which also led Vesta's seed round, returned to lead the Series B. Three of Vesta's own customers joined the round as investors: New American Funding, Pennymac, and nbkc bank. Citi Ventures, FirstKey Mortgage, and Navitas Capital came in as new backers, while existing investors Andreessen Horowitz, Zigg Capital, and Parker89, the venture arm of First American Financial, also participated.

Vesta was founded in 2020 by CEO Mike Yu and CTO Devon Yang. Yu told TechCrunch that demand for the product has exploded over the past year, with revenue up twelvefold year over year. Lenders currently using the platform, or in the process of migrating to it, originate more than $100 billion in mortgages annually, reported HousingWire.

AI Agent Swarms work inside the same loan file

The distinguishing idea behind Vesta's approach is that AI Agent Swarms operate inside the lender's core system of record rather than being bolted on top of legacy software. Yu described Vesta as the only loan origination system where people and AI agents work the same file in the same system, according to coverage of the company announcement.

In practice, a lender can choose to let the agents run document review, data validation, and cross-checks when new loans arrive. The agents record their decision-making and escalate anything unusual to a human reviewer. Many customers start with a human approving every agent action, then let the agents handle a growing share of loans on their own before expanding further, Yu explained to TechCrunch.

Some lenders are already using the agents to make mortgage underwriting decisions, although the companies themselves remain legally responsible for those calls. Yu emphasized that every action and the reasoning behind it is recorded for compliance and auditing purposes, a critical feature in a tightly regulated industry.

The design matters because mortgage software has historically been so rigid that only one person can work on a file at a time. According to an analysis published alongside the funding news, the cost of originating a single loan has roughly doubled since 2009, climbing from about $4,000 to $8,000, while the average mortgage still takes around 45 days to close. TechCrunch reported that a single mortgage can cost lenders around $11,000 to process under traditional workflows.

Pennymac's results set the pitch

The strongest evidence for the platform comes from one of its biggest customers. Pennymac, the second-largest mortgage lender in the United States, replaced its origination system with Vesta and reported a 25 percent decrease in its operational cost to originate, along with roughly 50 percent efficiency gains for loan officers, according to the company's funding announcement.

What makes that figure notable is that those gains arrived before the AI agents were even activated, according to reporting by citybiz. The modernization of the underlying data architecture alone moved the needle. The new funding is aimed at pushing the AI Agent Swarms further down the funnel, into tasks the system previously could not touch.

Yu pointed to recent model improvements as the unlock. He told TechCrunch that the breakthrough was Anthropic's Claude Sonnet 4.5, which the team found far better at following user-configured instructions over the longer time horizons that mortgage workflows require than previous model generations. A year earlier, the models simply were not reliable enough for the complex, multi-stage tasks involved in mortgage lending.

A crowded race for the mortgage stack

Vesta is not alone in chasing the mortgage industry's paperwork mountain. The startup competes against legacy incumbents like ICE Mortgage Technology, whose platforms were built decades before AI agents existed, and against fellow AI-native challengers such as Xpanse, all pitching AI Agent Swarms to the same lenders.

Yu argued that legacy vendors face a structural disadvantage. Because those platforms were never designed for AI agents, layering agents on top of them is extremely difficult, he told TechCrunch. Vesta's bet is that a system purpose-built for agent-human collaboration will win over systems retrofitted for it.

The financing also signals how enterprise AI Agent Swarms are converging with governance and audit requirements rather than avoiding them. Every automated underwriting step on the Vesta platform leaves a paper trail that auditors can inspect, which addresses one of the main objections enterprises have raised about agentic systems: the black-box problem.

What the raise means for the agent economy

The Vesta round fits a broader pattern of capital flowing into agentic software for regulated industries. This week alone has seen major agent funding rounds announced across the sector, reflecting investor confidence that AI Agent Swarms can take on real operational work rather than remaining glorified chatbots.

For Vesta, the stated priority is straightforward. Yu told TechCrunch that the immediate goal is earning the business of the rest of the mortgage industry, with new product lines including a personal assistant for mortgage issuers that can perform tasks and track workflows. The company also plans to expand beyond its current customer base once the mortgage beachhead is secured.

Whether agent-based origination becomes the industry standard will depend on how reliably AI Agent Swarms perform at scale, on regulatory comfort, and on the speed at which lenders retire their legacy systems. With three of its own customers writing checks into the round, Vesta has at least one constituency convinced that AI Agent Swarms belong in the mortgage business.