Figure’s AI Turns Abandoned HELOC Applications Into Funded Loans

While some companies are using artificial intelligence to eliminate the loan officer from the mortgage process, Figure is taking a different approach: using AI to make loan officers more productive, improve conversion rates, and help lenders capture revenue that might otherwise be lost to abandoned applications.

U.S. homeowners hold more than $35 trillion in home equity, yet fewer than half of home equity line of credit (HELOC) applications make it across the finish line. For lenders, this represents billions of dollars in potential lost revenue, not necessarily because borrowers lack interest, but because the process of getting an application from submission to funding can be operationally demanding. Loan officers are hired to build relationships, advise borrowers, generate business, and close loans, but they often spend hours helping applicants navigate basic administrative hurdles, such as taking a clear photograph of an identification document, locating tax returns, connecting financial accounts, or resetting passwords. These tasks can be time-consuming and frustrating for both parties, particularly when the application ultimately does not close. As a result, many lenders leave stalled applications untouched because the economics of pursuing them do not make sense.

We saw an opportunity to change that equation by applying AI to one of the mortgage industry’s most persistent challenges: loan abandonment. We work with approximately 500 partners, including independent mortgage banks, banks, credit unions, and fintechs, that use our technology to originate loans. Given the increasingly competitive lending environment and the importance of maximizing the value of every lead, we focused on developing an AI loan officer assistant that could help borrowers overcome application obstacles while allowing human originators to concentrate on higher-value activities.

The initial results are very encouraging. We’ve funded roughly $250 million in loans through this approach over just a handful of months. A cohort study we ran to evaluate results showed a 52 percent increase in total funded loans and a 37 percent increase in funded dollars, when comparing loan officer baselines to loan officers using an AI assistant. The conversion metrics also tell a clear story. Forty-four percent more applicants who received offers went through verification and funded loans, with a 29% increase in funded loan dollars per approval. 

While the average loan size fell 10% when using a loan officer assistant, the fact that funded dollars rose 37% clearly demonstrates the AI assistant’s proficiency in reaching smaller loans that would otherwise go overlooked. This is essentially found money for any company in mortgage, and it’s one of the things we’re most excited about. 

A key advantage: we  trained our  AI assistant using insights from our own origination process and substantial historical data. Since our founding in 2018, we’ve originated $30 billion in loans through our  platform – creating valuable experience in workflows and borrower interactions that influence successful loan completion. Our integrated technology and streamlined operating model also provide a foundation for the assistant to guide borrowers through the application process efficiently. Rather than simply generating more communications or increasing the volume of automated activity, the objective is to improve measurable business outcomes, particularly conversion and funded loan production.

In an industry increasingly inundated with AI vendors and promises of automation, generating more emails, calls, and chatbot conversations does not necessarily translate into better results. Our approach is intentionally on revenue outcomes, using AI to improve the quality and effectiveness of the lending process. The  assistant can patiently guide borrowers through repetitive tasks, answer questions, and help resolve issues that might otherwise require a loan officer’s time. An AI agent can spend 40 minutes helping someone locate an identification document without sacrificing the productivity of a salesperson who could be using that time to develop new business or advise a borrower with a complicated situation.

Importantly, our  initial results suggest that AI is most effective when it complements human expertise rather than attempting to replace it. In fact, the strongest performance in our rigorous testing came from combining the loan officer with the AI assistant. This reflects a broader pattern across professional financial services, where technology has made it possible to transact without an intermediary but has not eliminated consumer demand for personalized advice and guidance. Mortgage transactions are consequential financial decisions, and borrowers often value having a knowledgeable professional available to help them navigate their options. AI can handle repetitive administrative work, while loan officers focus on relationships, judgment, sales, and closing.

The division of responsibilities also has important implications for loan officer productivity and compensation. With AI handling more of the routine work associated with stalled applications, originators can spend more time engaging borrowers who are ready to move forward rather than leaving voicemails, repeatedly following up on missing documents, or troubleshooting technical issues. Our results indicate that loan officers can continue to originate business and earn commissions while focusing their attention on warmer, more qualified opportunities. For lenders, the potential benefits include higher conversion, improved production economics, and a better return on the leads they already generate.

Now, we’re extending this approach across its partner network: The AI assistant is now live with our wholesale partners, and we’re scoping and developing deployments for private-label and white-label partners. Interest from existing partners has accelerated the rollout, reflecting a growing demand for AI applications that produce tangible operational and financial results. 

For mortgage lenders and originators who remain skeptical of the current wave of AI adoption, we believe our strategy offers a practical alternative to sweeping promises of workforce replacement. The most valuable applications may not be those that attempt to automate the entire lending relationship, but those that remove friction from specific stages of the process, improve conversion, and allow experienced professionals to focus on the work they do best.

Ultimately, our AI loan officer assistant is built around a straightforward premise: abandoned applications represent unrealized business, and recovering that business should not require loan officers to spend hours on low-value administrative tasks. By combining the patience and availability of AI with the expertise and relationship-building skills of human originators, we’re excited to be helping lenders turn more applications into funded loans, improve the economics of home equity lending, and make loan officers shine. 

Alysse Guitar, VP, Marketing and Michael Tannenbaum, CEO, Figure