Blend debuts AI Copilot tool to improve LO workflow

Blend Labs is launching an artificial intelligence tool that can process mortgage application data for a loan officer and ease communication between originators and borrowers. 

The firm's Copilot platform can review a customer's financial profile, share pricing scenarios including closing costs, run the information through underwriting software and draft a pre-approval letter, said Nima Ghamsari, founder and head of Blend. The fintech is debuting the tool Tuesday afternoon at its customer conference, and will initially share it with clients through a waitlist. 

"This is truly something that is there to help them do the thing that they would like to do, which is to help the consumer, and to do a lot of the manual work," said Ghamsari. "No LO wants to be spending a bunch of time going to seven different systems to answer questions that are common. They want to focus on the hard questions."

Copilot, according to a demonstration, would be one of the more powerful AI chatbots among a field of emerging products from other fintechs and mortgage lenders. A prospective homebuyer in Blend's demo asks for a pre-approval letter through Blend's platform and Copilot produces it after an audit trail including processing through, for the purpose of the demo, Fannie Mae's Desktop Underwriter.

The LO can review and edit messages before they're sent to the consumer, and do the manual calculations if Copilot can't produce the desired result. Copilot also has SMS capabilities if a user wants to chat via text message, and questions only take a minute or so to process, Ghamsari said. 

The company, which already had a seven-year-old version of Copilot, developed the new platform in a few months given recent advancements in generative AI.

"The natural way that consumers and loan officers talk today is in a conversational format," said Ghamsari. "There's thousands of different ways [questions] can be asked and then an LO is smart enough to interpret that. There hasn't been a way for systems to really be able to interpret that, with this level of accuracy before."

The mortgage industry, while slower than its financial services peers in adopting technology, has latched on to the new wave of chatbots like ChatGPT. Outside of a few firms' own generative AI models, LOs this summer said they've used ChatGPT to improve their web searches, write marketing copy, and aid with editing borrower correspondences. 

While lenders are using AI to sift through reams of data, the tech is far from replacing traditional underwriting and fulfillment, experts say. Regarding Blend's Copilot, Ghamsari emphasized there is always an LO between the consumer and the tech.

"We want to make sure that the answers going to the consumer are right," he said. "If you're a lender, you want to make sure from a compliance perspective that you have the appropriate level of certainty around the responses that go out."

The product also deploys at a difficult time for the industry amid sky-high mortgage rates and decades-low application activity. Ghamsari said Blend's customers are using this time to prepare for the next wave of originations should rates become more amenable to borrowers. 

The Bay Area-based Blend is coming off a $36.7 million net loss in the second quarter, although it's been progressively shrinking its deficit. It also announced in August a layoff of 150 employees. The fintech is projecting profitability in 2024.