Sela announced $21 million in Seed and Series A funding on 22 September 2026, saying its voice AI agents help loan officers originate more than $1 billion in new mortgages each month, according to its funding announcement.
Key points
- Costanoa led the funding, with participation from Emergence Capital, Sela says.
- Sela says six of the ten largest independent mortgage banks use its agents.
- The company reports annualised run-rate revenue above $10 million after 18 months.
- In one lender test, Sela says its agents raised the lead-to-lock rate by 9%.
Costanoa leads Sela’s $21 million funding
The $21 million covers two funding rounds rather than a single new round. Sela named Costanoa as lead investor and Emergence Capital as a participant. It said the money would support hiring and further development of agents for other stages of consumer finance.
Sela says it passed $10 million in annualised run-rate revenue in 18 months. That is the company’s revenue measure. Its separate figure of more than $1 billion a month describes new mortgages that loan officers originate with help from its agents.
David Cheng, a partner at Costanoa, said the investor’s decision centred on the prospect of more funded loans for lenders. He identified conversion and cost per funded loan as the two measures under pressure in a lender’s profit and loss account. Cheng said six of the ten largest independent mortgage banks had put Sela into production in under two years.
The company was founded by Nate Becker, previously a co-founder of VoiceOps and a data scientist at LinkedIn, and Vahe Tshitoyan, formerly a senior machine learning engineer and tech lead at Google. Sela is based in San Francisco and builds agents for consumer lending.
Sela tests conversion against lenders’ existing processes
Sela says its agents speak with prospective borrowers, answer questions and bring a loan officer into the conversation when needed. The company designs and operates the agents for each lender, and says it measures borrowers reached, productive conversations and completed loans. Its stated measure of success is tied to lending outcomes rather than the number of minutes an agent spends on calls.
Large lenders typically test the agents against their existing systems before a wider rollout, Sela says. In an A/B test involving more than 10,000 prospective borrowers, the company reports a 9% increase in the rate at which leads reached a mortgage-rate lock. It says the lender’s profit in that test was more than 40% higher than under its existing process.
A separate A/B test compared Sela with another voice AI product. Across more than 7,000 leads at what Sela describes as a top-five mortgage servicer, the company says its agents performed 41% better on a lead-to-lock basis.
Those performance figures come from Sela’s own reporting. Forkast described the company’s metrics as entirely self-reported on 23 September and said long-term, third-party audits would be needed to validate them.
Sela plans growth beyond 17 employees
Becker, Sela’s co-founder and chief executive, said the company draws on sales behaviour learned from tens of millions of calls. He said the agents would take on more of the mortgage sales process over the 12 months following the announcement, extending their role beyond the borrower conversations they already handle.
Sela said on 22 September that it had 17 full-time employees and planned to grow to 50 over the following year, hiring in product, engineering and go-to-market roles.
