Blend Labs, Inc. 2026 Q2 Earnings Call

Please go ahead. Good afternoon, and welcome to Blend's Financial Results Conference Call for the second quarter of 2026.

I'm Meg Nunnally, Blend's Head of Investor Relations. Joining me today is Nima Ghamsari, our Co-founder and Head of Blend, and Jason Ream, our Head of Finance and Administration. Before we start today's call, I'd like to note that we will refer to certain non-GAAP measures, which are reconciled to GAAP measures in today's earnings release and in the appendix of our supplemental slides. Non-GAAP measures are not intended to be a substitute for GAAP results. Unless otherwise stated, all financial measures we'll discuss today, including our profitability, refer to non-GAAP.

Certain statements made during today's conference call regarding Blend and its operations, in particular our guidance for the third and fourth quarter of 2026, other commentary regarding 2026, and our expectations about markets, our strategic investments, product development plans, and operational targets may be considered forward-looking statements under federal securities law. We caution you that forward-looking statements involve substantial risks and uncertainties, and a number of factors, many of which are beyond the company's control, could cause actual results, events, or circumstances to differ materially from those described in these statements. Please see the risk factors we've identified in our most recent 10-Qs, our 10-K for the fiscal year 2025, and other SEC filings. We are not undertaking any commitment to update these statements if conditions change, except as required by law.

The financial information presented on this call is based on continuing operations, and prior periods have been recast to exclude operations that are now discontinued. We'll be providing a copy of our prepared remarks on our website by the conclusion of today's call, and an audio replay will also be available soon after the call. I'll now turn the call over to Nima.

Thanks, Meg. Welcome everyone. The second quarter was another disciplined, profitable quarter for Blend. Revenue came in near the high end of our guidance range, non-GAAP operating income came in above the high end. Jason will take you through all the financial details in a few minutes. Today, I want to spend my time on the two pillars of our strategy. Autopilot, the agents we build for our customers, and Blend 3.0, the agents we are building inside Blend to help us do our work to serve our customers faster, better and cheaper. Let me start with Autopilot. The big news for us last quarter was that Autopilot became commercially available on July 1st. If you're following along on the webcast, I'd invite you to advance the slide with the words "Our AI strategy, guiding our customers" at the top.

This is the first of two slides that I'll reference today. The charts on these slides are also available in the supplemental slides on our investor relations website. The chart shows cumulative loans processed by Autopilot from a standing start in February to more than 45,000 today. It's still compounding. The curve is the proof behind everything I'm about to tell you. During this four-month window, more than 65 lenders activated Autopilot. They stress-tested, surfaced the hard edge cases, and shaped what we shipped. The preview data backs up why this matters. We're starting to see evidence Autopilot is driving faster clearance times, higher conversion rates, and potentially reducing fulfillment costs. Based on our preliminary data of these loans that have gone through our system, our customers are seeing a 10%-15% improvement in pull-through rates and two to four days of cycle time improvement.

Furthermore, we estimate that Autopilot is automating four and a half hours of loan fulfillment tasks on average per loan. This is huge. Our customers are just beginning to grasp the potential. Now that we are commercial as of July 1st, six lenders have already signed contracts that include Autopilot, including Onity, which is one of the largest mortgage servicers in the nation, which is also building its own experiences on top of Autopilot through our Autopilot MCP server. On monetization, we're executing the plan we described in May, with customers signing flat-fee, one-year contracts for full access. There's going to be a base level of intelligence built into our workflows, the paid tiers are where the full product lives, what we call our underwriting intelligence, where Autopilot is reading documents, running calculations, reconciling against guidelines, and driving the full loan file forward.

Over time, our intent remains to move the paid tiers to a per-funded loan model, just like the rest of our Mortgage Suite. When seat-based pricing, which we don't think survives the agentic world, when software does the work, you can't charge by the person. It doesn't make any sense. Others are also moving the consumption models to get paid for the activity their AI generates. We made a different choice, which is we get paid on success. We get paid on the outcome, not for the tasks along the way, but for the outcome. If Autopilot does 10 times the work on a file that never closes, our customers shouldn't pay 10 times more for it, under our model, they won't. Our revenue scales with our customers' success, that alignment is the business model we've always had.

Agentic AI just makes it more valuable and more scalable. Since I get asked about the competitive landscape constantly, let me be direct about it. What gives Blend the right to win? Our answer comes down to four advantages that are hard to replicate. First, where we sit. Blend is the borrower's first point of contact, and because we're there, we see the problems, and we fix the problems as they come along. More than half of borrowers apply outside of business hours, and over 90% of people who complete an application do so within 24 hours of starting. Autopilot catches friction the moment it happens, at 10:00 P.M. on a Saturday, not Monday morning, at the exact moment the borrower intent is the highest and the engagement is the highest. Next, our data. The question behind every AI question I get is, in a world where anybody can call a frontier model like Claude, is a software like ours replaceable?

We believe the answer is no, and the reason is our data. Models are converging, and everyone has access to the same models, including us. What isn't a commodity is what it takes to get the model to perform on a mortgage that has tons of context, tons of loan documents, guidelines, and lots of things that have to be taken into account to make the right next action on the loan. That's a combination of 15 years of experience and tens of millions of loan application data processing through our platform, which you can't buy. You can't synthesize 15 years of real borrower behavior. That compounds. I actually see this in our early Autopilot benchmarks against a typical Claude plus skills, and we see that Autopilot performs better in those tests, and much better and much cheaper, about one-third the cost.

Third, the harness. Autopilot is not a wrapper around a generic model. It runs inside the infrastructure, the data layer, the integrations, the compliance architecture, orchestrating what the agent sees, what tools it can use, which guidelines to apply to a specific loan in front of it, and what happens when it isn't sure. It hands the file back to the loan team. People stay in control of the decisions that matter, and this harness is part of what makes Autopilot more accurate and cheaper than a generic harness. Last, our relationships. We've spent 15 years building alongside lenders, and Autopilot was built the same way, with lenders, for lenders.

With Autopilot MCP, we've opened up our infrastructure so customers and other technology providers can build on top of us rather than around us. When you step back, this is the bigger thesis. Blend is and can be the agentic infrastructure for relationship banking. The original promise of banking was a relationship, a lender who got to know you and could make a call based on more than just a credit score. That promise didn't disappear because bankers stopped caring. It disappeared because there's so much process, and there's so much manual work and manual effort that has to go in to every single loan.

When Autopilot is handling that grunt work on the loans and helping the people who historically did that focus on the customer, when Autopilot's handling those conditions, the follow-ups, the questions that come in at 2:00 A.M., and the loan officer bringing it back to the capacity of serving the customer. I started Blend with my co-founders back in 2012 with a simple thesis. The mortgage process should drive itself, not because humans aren't needed, because the right technology can handle everything that does not require a human. Autopilot is finally that thesis arriving.