40 Days, $11,000 a Loan – AI Is Working the Whole Pipeline

When Pennymac and New American Funding led Vesta's $30M Series B in October 2026, they were not making a financial bet – they were making an operational one. Both are active customers. Both cited the same metric: a 25% reduction in cost-to-originate. Their check was the endorsement.

Vesta builds AI agent infrastructure for mortgage lenders. The problem it is solving is structural: the average US mortgage takes roughly 40 days to close and costs the lender approximately $11,000 to originate. Most of that cost is not capital – it is coordination. Documents get requested, re-requested, routed to the wrong processor, and manually checked against compliance rules that change quarterly. Vesta replaces that coordination layer with agents.

The platform handles document ingestion, condition clearing, compliance verification, and pipeline management. It connects to more than 100 prebuilt integrations – LOS systems, POS platforms, pricing engines – which means a lender does not need to rebuild its stack to adopt it. The agents work within existing infrastructure, not in place of it.

The growth numbers are unusual for a six-year-old company. Revenue grew 12x year-over-year. AI agent usage on the platform grew 988%. Vesta's CEO notes the company still holds less than 5% of the addressable market, which in mortgage origination is enormous: US lenders close roughly 4 million purchase loans per year. At $11,000 per loan, a 10% market share would represent $4.4B in annual cost that the platform is actively compressing.

The round brings total funding to $85M. The capital is earmarked for headcount – the CEO used the phrase 'staff up' explicitly – and for deepening the agent capabilities across the origination pipeline. The company is not pivoting. It is scaling a model that already works at named enterprise accounts.

For builders watching this space: the opportunity is not in replicating Vesta. It is in identifying analogous workflows in adjacent regulated industries. Insurance underwriting, commercial real estate title work, auto lending decisioning – each has a similar profile: high per-transaction cost, heavy document burden, compliance overhead, and a fragmented software stack. The agent-orchestration architecture that works in mortgage does not need to be rebuilt from scratch for those markets. It needs to be adapted.