Vesta, an AI-native software startup focused on helping lenders originate mortgages, announced on Thursday that it has raised $30 million in a funding round. The investment was led by Conversion Capital, with additional contributions from three of Vesta's customers, Pennymac and New American Funding, as well as Citi Ventures and Andreessen Horowitz.
The startup employs AI agents to automate various stages of the loan origination process. According to the founders, this approach helps lenders decrease the time and expenses associated with processing mortgages. Mike Yu, Vesta's CEO and co-founder, indicated that the timing for this funding round was opportune due to a significant increase in product demand over the past year.
Yu reported that the company's revenue has grown 12 times year over year, and Vesta has assisted lenders in originating over $100 billion in loans annually. The company has now accumulated $85 million in total funding. Yu stated that Vesta's current market share is under 5%, and the new capital will be used to expand staff, capture more market share, and invest in new product lines.
Among the planned new offerings is a personal assistant designed for mortgage issuers, intended to help with tasks and workflow tracking. Yu noted that closing a mortgage in the U.S. typically takes around 40 days and costs approximately $11,000 per loan, with much of this cost attributed to human labour and review bottlenecks. Vesta's system allows humans to deploy multiple AI agents to accelerate task completion, with customers determining which tasks to assign. Some lenders are reportedly using Vesta AI agents for mortgage underwriting decisions.
Yu clarified that companies remain responsible for underwriting decisions, regardless of the software or AI agents used. He added that all actions and reasoning behind AI decisions are recorded for compliance and auditing purposes. This increased autonomy is attributed to recent advancements in AI models, specifically mentioning Claude Sonnet 4.5 as a significant breakthrough for adhering to user instructions over necessary timeframes.