HomeLight recently launched EVA, an artificial intelligence-powered escrow agent designed to automate most tasks needed to complete residential real estate transactions.
The company also announced $40 million in new debt financing from funds managed by BlackRock to expand its platform nationally.
Drew Wooher
A typical escrow requires more than 120 separate tasks that are time-consuming, difficult to coordinate, and error-prone. HomeLight says EVA automates much of the process, from initiating orders and retrieving HOA documents to communicating with lenders and transferring funds.
“When you look at the intersection of AI and rights and escrow, it’s really amazing,” HomeLight founder and CEO Drew Wooher told Inman. “We feel it’s practically a foregone conclusion that this industry will be fundamentally different in five to 10 years, because there’s a lot of repetitive back-office work involved in closing escrow.”
From frustration to “excitement when the lights go out”
The product will take approximately 19 months to develop, with the first automated workflows expected to go live in early 2025. Wooher said accuracy rose quickly from 25 to 30 percent at launch to the 80 to 90 percent range, then plateaued there for about six months.
“As of December 2025, I was honestly frustrated with the project,” he said. “I didn’t even know if this was a problem that could be completely solved.”
Apparently so. Wooher said that by the first quarter of 2026, EVA will achieve virtually 100% accuracy, first in one workflow, then another, and then again.
“From February to early May of this year, all four major workflows reached 100% accuracy,” Uher said. “We’re really excited about this. We’ve now automated about 25 percent of the surface area of the escrow process, and it gives us confidence that we have the prospects to automate most of it.”
This framework requires caution. Wooher defines 100% as not zero uncertainty, but zero false confidence.
“When I say 100 percent accuracy, it’s okay if EVA runs into a really weird edge case and says, ‘I don’t know how to deal with this, it’s escalating.’ It still counts as 100%,” Wooher said. “What I won’t tolerate is an EVA saying, ‘I think I know how to do it,’ and then doing it the wrong way or making a mistake. With title and escrow, you’re dealing with people’s money, often their life savings, so the stakes are very high.”
Inside the EVA workflow
EVA stands for “Escrow Virtual Assistant”. Wooher described it as a virtual workforce, a collection of different AI agents. “The secret sauce is that we’ve built over 80 different tools that EVA can use to access the outside world and get real work done,” he said.
Wooher was quick to note that EVA is not a “chatbot experience.”
“What we’re talking about is an AI agent, or a collection of AI agents, that can actually go out and do things,” Uher said.
For example, EVA monitors the escrow agent’s inbox, waits for open order requests to arrive, opens emails and PDFs, and reviews the 20+ critical pieces of information needed to open the order. Wooher said “she” then goes into the records system and enters all of that information.
“If something is missing, she contacts the agent directly: ‘I don’t have the buyer’s phone number, do you know it?’ — waits for a response and adds it to the file,” Uher said.
Once the order is open, we will email the order number to the agent, begin the title search process with the title vendor, begin the HOA process if associated with the file, send inquiry forms to the buyer and seller, repeat any questions, and notify the buyer where to send the earnest money.
Regarding the HOA process, Wooher said certain states have websites with databases of HOA documents. Texas has two or three major sites that contain most of the state’s HOAs.
EVA can also search these sites, retrieve relevant documents, and purchase documents on behalf of the company using HomeLight’s credit card. She also has access to hundreds of county tax and assessor offices and various regulatory agencies.
The EVA product roadmap still includes advanced title analysis, curation efforts, signing schedules, and more. But EVA also automated lender authority requests, notary QC, post-closing preparation, and several other parts of the flow.
competition window
HomeLight entered title and escrow in 2019, predating the advent of the AI tools that are now making this automation more manageable. Wooher admitted that the pace was slow in the early years. All of the relevant information for a typical escrow was stored on paper, and a more robust AI model was needed to reliably extract it, but that model wouldn’t arrive until 2023 or 2024.
Opportunities for competition are really opening up, he argues.
“If you look at the title and escrow industry, it’s probably in the top 1 percent of all industries where AI can make a real, meaningful difference, and no one is actually doing anything about it,” Wooher said.
Wooher continued that “large established companies” are generally not technologists.
“Due to the state of the capital markets, new startups are currently not getting funding,” he said. “And most of the existing title and escrow players have been sold or consolidated. So no one is innovating in this space right now.”
He pointed out that some software companies are building systems for rights and escrow, but they’re building it for all of their customers because they have thousands of customers, each with their own workflow.
“We are an AI-first title and escrow agency with one workflow: a workflow, and we are building technology to ensure escrows close 100 percent on time and with 100 percent accuracy,” Uher said.
HomeLight is currently the only customer of its technology. The company has not yet licensed EVA to other title agencies.
bigger story
Victor Lund, managing partner at WAV Group and CEO of RE Technology, believes the launch is legitimately significant, but says the headlines may be understating what’s really at stake.
Victor Land
“The data coming out of escrow is extremely valuable,” Rand told Inman.
He pointed to the market caps of data companies like Cotality and Black Knight as a framework for what HomeLight could ultimately access if it scales.
“If we could get all this information at scale, we could license it to those companies or directly to capital markets players,” Rand said. “Brokers and agents aren’t even thinking about the value of that data. They’re just processing trades.”
Lund also warned about security concerns. “If AI is processing wire transfer instructions, there’s no human being whose screen can be compromised,” he said, noting that wire fraud is one of the most common and costly forms of real estate fraud.
It’s unclear whether EVA is actually safer than the current situation, and “none of us have seen it yet,” he cautioned.
What’s even more unknown, Lund says, is the market development strategy.
“How do you sell escrow?” Lund said. “We have to retrain real estate agents and brokers. And first of all, brokers want to use their own escrow companies, so there’s built-in competition. Market development is the hardest thing we don’t know.”
He said HomeLight’s path forward continues through its 2020 acquisition of Disclosures.io, which became the standard listing and offer management platform for Northern California transactions.
“The strategy is to go to all of Disclosures.io’s customers and say, ‘Would you like to handle your escrow for us?'” Lund says. “They can offer lower prices and better service. And as long as there are no kickbacks or rev shares, it’s RESPA compliant. It’s just a better deal.”
“That’s all there is to profit.”
Wooher is outspoken about the impact on labor, but he draws a line.
“Right now, 80 to 90 percent of the work we do in escrow is repetitive back-office work that frankly no one wants to do,” he said. “AI can handle all of that.”
His framing for the humans remaining in the process: Own one of the two tracks. The first is the role of the agent-facing service. The role is responsible for agents and clients when things go awry, what Wooher called a “choking situation.”
The second are escalations and edge cases, the truly bespoke corners of transactions where human judgment is still required.
Lund put it more bluntly. A good escrow officer will probably handle 20 closings at a time. If AI handles the checklist work and humans handle exception handling, the same executive could potentially manage 100. “It just pays dividends,” he said.
Email Nick Pipitone
