Vesta secures $30M to expand AI mortgage origination platform

Vesta raises new capital for AI-driven mortgage workflows

Vesta, an AI-native software company for mortgage lenders, has raised $30 million in a financing round led by Conversion Capital. Pennymac and New American Funding, both Vesta customers, joined Citi Ventures and Andreessen Horowitz as investors. The startup, founded in 2020 by Mike Yu and Devon Yang, says it has now raised $85 million in total funding.

The company builds AI agents for tasks across mortgage loan origination. Vesta says its software helps lenders reduce the time and cost involved in processing loans, and that its platform supports more than $100 billion in annual loan originations. Chief executive Mike Yu said revenue increased 12-fold year over year as demand for the product rose during the past year.

Agents are introduced with lender-defined controls

Mortgage closing in the United States takes roughly 40 days and costs about $11,000 per loan, Vesta said. Yu attributed much of that expense to human labour, with delays often arising while a reviewer waits to examine a loan. The company’s model is to let lender staff decide which tasks an agent should perform across the multi-stage process.

Customers can begin by having a person approve an agent’s work, then allow the agent to handle a portion of loans independently before expanding its remit. Yu said some lenders are using Vesta agents to make mortgage underwriting decisions. He also stressed that lenders remain responsible for underwriting decisions regardless of the software or AI agents they use.

For compliance and audits, Vesta records the actions taken by its agents and the reasoning behind a decision. That record is particularly relevant in a regulated workflow where lenders must be able to review how automated systems were used, even as the level of agent autonomy increases.

Model capability underpins the product expansion

Yu said earlier AI models were not capable enough for the complex, multi-stage work involved in mortgage lending. Vesta had initially concentrated on data architecture that could support advanced automation tools. He identified Anthropic’s Claude Sonnet 4.5 as a key improvement because it adhered more reliably to user-configured instructions across the time horizons Vesta requires.

The funding will support hiring, market expansion and new product lines, including a personal assistant for mortgage issuers that can perform tasks and track workflows. Yu said Vesta remains below 5% market share and aims to win business across the broader mortgage industry.

Competitive pressure on mortgage technology platforms

Vesta competes with established mortgage systems such as ICE Mortgage Technology as well as AI-native providers including Xpanse. Yu argued that legacy platforms were not built for AI agents, making it difficult to add agent-based capabilities on top of their existing systems.

For mortgage businesses, the practical implication is to define the tasks, approval stages and audit requirements before widening agent autonomy. Vesta’s approach places automation inside lender-selected controls while leaving responsibility for underwriting with the lender.