Vesta Raises $30M as Mortgage Lenders Let Agents Underwrite
Vesta, a loan origination platform founded in 2020 by two former Blend employees, announced a $30 million round on Thursday led by Conversion Capital, with Citi Ventures, Andreessen Horowitz and three of its own customers, including Pennymac and New American Funding, investing. Total raised is $85 million. CEO Mike Yu told TechCrunch revenue is up 12x year over year and that demand "exploded in the last year," and the plan is to staff up, take share and ship new product lines while the company is still under 5% of the market.
The product is a swarm of agents inside the mortgage process. A US mortgage takes about 40 days and $11,000 to close, and Yu's claim is that most of that cost is labor and most of the delay is waiting for a human to get to the file. Lenders choose which tasks to hand the agents. The typical path is to start with a person approving every action, then let the agent handle a share of loans on its own, then widen it. Some lenders are now letting Vesta agents make underwriting decisions outright. Every action and the reasoning behind it is logged for compliance and audit, and the lender stays legally responsible for the decision regardless of what software produced it.
The detail worth pulling out is the model line. Yu says the breakthrough was Claude Sonnet 4.5, which was "much better at adhering to user-configured instructions over the time horizons we need than previous generations." Before that, the company spent its time on data architecture so that it would be ready when a model could hold instructions across a multi-stage, weeks-long workflow. That is a precise statement of what unlocked agent adoption in a regulated vertical: not raw intelligence, but instruction adherence over long horizons.
A mortgage is about the most consequential, most regulated decision an agent can be put near, and lenders are already past the "human approves everything" stage with real customers' loans. The logged-reasoning requirement is the compliance wrapper that made that possible. Expect the same shape in insurance and lending generally: agent decides, trace is kept, institution signs.
Links: techcrunch.com/2026/10/08/vesta-raises-30m-as-lenders-adopt-ai-agents/, usevesta.com
← Back to all articles
The product is a swarm of agents inside the mortgage process. A US mortgage takes about 40 days and $11,000 to close, and Yu's claim is that most of that cost is labor and most of the delay is waiting for a human to get to the file. Lenders choose which tasks to hand the agents. The typical path is to start with a person approving every action, then let the agent handle a share of loans on its own, then widen it. Some lenders are now letting Vesta agents make underwriting decisions outright. Every action and the reasoning behind it is logged for compliance and audit, and the lender stays legally responsible for the decision regardless of what software produced it.
The detail worth pulling out is the model line. Yu says the breakthrough was Claude Sonnet 4.5, which was "much better at adhering to user-configured instructions over the time horizons we need than previous generations." Before that, the company spent its time on data architecture so that it would be ready when a model could hold instructions across a multi-stage, weeks-long workflow. That is a precise statement of what unlocked agent adoption in a regulated vertical: not raw intelligence, but instruction adherence over long horizons.
A mortgage is about the most consequential, most regulated decision an agent can be put near, and lenders are already past the "human approves everything" stage with real customers' loans. The logged-reasoning requirement is the compliance wrapper that made that possible. Expect the same shape in insurance and lending generally: agent decides, trace is kept, institution signs.
Links: techcrunch.com/2026/10/08/vesta-raises-30m-as-lenders-adopt-ai-agents/, usevesta.com
Comments