Vesta raises $30M as AI agent swarms move into mortgage underwriting

Vesta lands $30M for AI-driven mortgage origination

Vesta, an AI-native software startup that helps lenders originate mortgages, has raised $30 million in a round led by Conversion Capital, TechCrunch reported on Thursday. The company uses fleets of AI agents to automate much of the loan origination workflow, and its founders say the approach cuts both the time and the cost of processing home loans.

According to TechCrunch, the round drew participation from three of Vesta's customers, including Pennymac and New American Funding, as well as Citi Ventures and Andreessen Horowitz. The company has now raised $85 million in total since Mike Yu and Devon Yang founded it in 2020.

Yu, who serves as CEO, told TechCrunch that demand has surged over the past year, with revenue up 12 times year over year. He said Vesta now helps lenders originate more than $100 billion in loans annually. Even so, he described the company as holding under 5% of the market, and said the new capital will go toward hiring, expanding market share and building new product lines.

A graduated path to agent autonomy

Those new products include a personal assistant for mortgage issuers that can carry out tasks and track workflows. The backdrop, as Yu laid out for TechCrunch, is a US mortgage process that takes roughly 40 days to close and costs around $11,000 per loan. Most of that cost is human labor, and a significant bottleneck is simply waiting for a person to review a loan file.

Vesta's answer is to let human operators dispatch a swarm of agents to work through tasks in parallel. Customers choose which tasks to delegate, and adoption follows a pattern of increasing trust: lenders typically begin with a person approving each agent's output, then allow the agent to handle a share of loans without supervision, then expand from there. Yu said some lenders now use Vesta's agents to make mortgage underwriting decisions.

That autonomy comes with guardrails. Yu stressed to TechCrunch that lenders remain responsible for underwriting decisions regardless of what software or agents they use, and that every action and the reasoning behind a decision is recorded to support compliance and allow AI decisions to be audited.

Why the timing worked

The shift toward more autonomous agents reflects how far AI models have come in the past year, Yu told TechCrunch. Earlier generations were not reliable enough for the complex, multi-stage tasks involved in mortgage lending, so Vesta initially concentrated on building the right data architecture to support automation. He pointed to Anthropic's Claude Sonnet 4.5 as the breakthrough, saying it adhered far better to user-configured instructions over the long time horizons these workflows demand.

A three-way competitive fight

Vesta is competing on two fronts, according to TechCrunch: against traditional mortgage software providers such as ICE Mortgage Technology, and against fellow AI-native companies like Xpanse that are also automating lending. Yu argued that legacy platforms were never designed for AI agents and that bolting agents on top of them is very difficult. Looking ahead, he said the priority is winning the business of the rest of the mortgage industry, after which the company will follow wherever its customers take it.

Why it matters

Mortgage origination is one of the most heavily regulated, document-intensive processes in consumer finance, and it has historically resisted automation beyond rules-based scripting. Vesta's growth suggests that agentic AI is now crossing into decisions with real legal and financial consequences, not just drafting emails or writing code.

Two details stand out. First, lenders themselves invested in the round, which signals that customers see the product as strategically important rather than experimental. Second, agents are being used in underwriting, an area where regulators and courts expect explainability; Vesta's logged reasoning trail is an early template for how autonomous systems might operate inside compliance regimes built for human decision-makers.

The competitive dynamic also matters beyond mortgages. If Yu is right that retrofitting agents onto legacy systems is very hard, AI-native entrants may have a structural advantage in other regulated back-office markets, such as insurance and commercial lending. And with Vesta at under 5% market share but already touching over $100 billion in annual loan volume, the round is a bet that agent-driven origination becomes the default rather than the exception.