Sela raises $21M for mortgage voice AI agents

Sela, a voice AI system for mortgage sales, announced $21 million in total Seed and Series A funding led by Costanoa with Emergence Capital, saying its agents help loan officers originate more than $1 billion in new mortgages per month and that the company crossed $10 million in annualized revenue in 18 months.

FINANCE desk — voice agents are already sitting in the mortgage conversion funnel at large independent banks, paid on funded loans not minutes.

What the company says the agents do. A voice AI agent, here, is software that talks with a person on the phone. Sela says its agents answer complex questions, build rapport, educate borrowers, and bring in a loan officer when necessary. A loan officer is the licensed person who can take a mortgage application forward. The release says the agents are built to maximize conversion, and that Sela designs, runs, and keeps improving them for each lender. Conversion, here, means a higher share of people who start as a lead and move toward a loan. The company says it measures success in borrowers reached, productive conversations held, and loans closed, not minutes used. File that pitch as the company’s. This desk did not listen to a call.

The scale lines are the company’s, and they are not an audit. The release says Sela’s AI agents help loan officers originate more than $1 billion in new mortgages per month. Originate, here, means start a new mortgage that is on its way to being funded. The same sentence says the company crossed $10 million in annualized run-rate revenue in 18 months. Annualized run-rate revenue, often shortened to ARR, is the yearly value of current sales if the recent pace holds. It is not audited profit, and it is not cash already booked for the whole year. The subhead says 6 of the 10 largest independent mortgage banks use Sela. An independent mortgage bank, here, is a lender that is not a deposit-taking commercial bank. David Cheng, a partner at Costanoa, says six of the ten largest independent mortgage banks put Sela in production in under two years. Six of ten, and under two years, are those sentences. This desk did not see the bank list or the books.

The test results are company claims. An A/B test is a side-by-side trial: one group gets the new process, another keeps the old one. Lead-to-lock is the share of leads that reach a rate lock. A lead is a person who might want a loan. A lock is the moment the interest rate on that loan is set. In one test of more than 10,000 prospective borrowers, the release says Sela produced a 9% increase in lead-to-lock rates, and that the lender’s profit was more than 40% higher than with the existing process. In a second test, with a top-five mortgage servicer, the release says Sela beat another voice AI by 41% on a lead-to-lock basis, across more than 7,000 leads. A mortgage servicer collects payments on loans that already exist. The release does not name the servicer or the other voice AI. File the 9%, the more-than-40% profit line, and the 41% as Sela’s. This desk did not see the test files.

Who founded it, and how big the team is. The release says Nate Becker, co-founder and chief executive, was a co-founder of VoiceOps and a data scientist at LinkedIn. Vahe Tshitoyan was a senior machine learning engineer and a tech lead at Google. The company has 17 full-time employees. It plans to grow to 50 over the next year, hiring in product, engineering, and go-to-market. Go-to-market means the people who sell and set the product up with lenders. The release says the funding will grow the team and keep developing agents that guide borrowers through the entire consumer finance journey. That wider journey is a plan. This filing is the mortgage sales product the release describes today. The about box puts headquarters in San Francisco. The company site is sela.io.

Two quotes, as color only. Nate Becker said that at his last company the strongest salespeople outperformed the average by three or four times, and that Sela’s agents take the best-performing sales behaviors, learned across tens of millions of calls, and use them on every interaction. He said that in the next 12 months the agents will run more of the mortgage sales process. David Cheng said every lender’s profit-and-loss statement comes down to conversion and cost per funded loan, and that what convinced Costanoa to lead was that the founders sell the ability to do more funded loans. A profit-and-loss statement is the tally of money in and money out. File the names, the titles, and those sentences as the release’s. A quote is not a customer list, and it is not a price for the company.

Do not fill in a price the release left blank. The PR Newswire page does not print a valuation. A valuation would be the price the round puts on the whole company. Do not add one. Do not split the $21 million into a seed check and a Series A check. The company homepage at sela.io prints different figures from this release. It says 15 million or more monthly calls, $6 billion or more in originations, 6 of the top 10 largest U.S. mortgage lenders, and 3 of the top 5 aggregator lead buyers. An aggregator, here, is a firm that buys leads and resells them. Those homepage lines do not say per month for the $6 billion, and they say mortgage lenders, not independent mortgage banks. Do not fold $6 billion into the release’s more-than-$1-billion-a-month line. Do not treat the two “6 of 10” sentences as the same list. This desk did not reconcile them.

Plain English for the rest of the card: Seed = an early funding round. Series A = the next growth round. $21 million is the total the release prints across both. voice AI agent = software that talks with a borrower on the phone. loan officer = the licensed person who can move a mortgage application forward. lead = a person who might want a loan. rate lock = the moment the interest rate on that loan is set. lead-to-lock = the share of leads that reach a lock. originate = start a new mortgage. ARR = annualized run-rate revenue, the yearly value if the recent sales pace holds. $10 million in 18 months is the company’s figure. independent mortgage bank = a lender that is not a deposit-taking bank. mortgage servicer = a company that collects payments on loans that already exist. This filing is the 22 Sep announcement. It is not a priced public listing.

PRIMARY here: Sela’s 22 Sep 2026 release on PR Newswire, “Sela's AI Agents Help Originate $1B+ in Loans per Month, Company Raises $21M,” stamped 10:00 ET and datelined San Francisco — Tier A PRIMARY, the company’s own record. The company homepage is a product page, not a second newsroom. The $21 million total across Seed and Series A, the Costanoa lead, Emergence Capital, the more-than-$1-billion-a-month origination line, the $10 million annualized run-rate in 18 months, the 6 of 10 largest independent mortgage banks, the two A/B tests, the Becker and Tshitoyan biographies, the 17 full-time employees, the plan to grow to 50, and the use of funds are the release’s. The homepage’s 15 million or more monthly calls, the $6 billion or more originations line, the 6 of the top 10 U.S. mortgage lenders line, and the 3 of the top 5 aggregator line are the site’s, and they are not the same sentences as the release. NOT claimed: a valuation, a split of the $21 million, the names of the six banks, the name of the other voice AI, that this desk audited originations or the A/B files, a stock tip, or investment advice. Distinct from the already-filed chamelio-26m, snorkel-350m, and spott-21m.