Kastle Raises $24M for AI Loan Operations

The expensive part of lending automation begins after a borrower says something that requires action. A payment, escrow question, loan inquiry or collections call can touch several systems, create a compliance obligation and end in a case that still needs human judgment.

Kastle has raised a $24M Series A to build more capacity into that operating layer. Insight Partners led the September 17, 2026 financing, with Y Combinator and Commerce Ventures returning. Fifth Wall and a group of unnamed founders and financial-services executives joined as new investors.

The round matters because Kastle is asking banks and lenders to adopt AI without first replacing the core systems already carrying borrower, payment and servicing records. Its agents are designed to work across those systems, complete approved workflows and return difficult cases to people. The company is betting that the practical path to AI-native lending runs through existing infrastructure, not around it.

What Kastle's $24M Series A Funds

Kastle said it will use the Series A to expand engineering, product and go-to-market teams, deepen platform capabilities and accelerate deployments with banks and other financial institutions across North America. The company did not disclose a valuation, round terms, ownership changes, board appointments or individual investor check sizes.

The financing follows a prior $2.3M raise disclosed by co-founder and CEO Rishi Choudhary. Publicly announced round arithmetic therefore reaches at least $26.3M, although Kastle has not presented that figure as an official total and the composition of its earlier capital is not fully reconciled.

Insight Partners led the new round, and Managing Director Rebecca Liu-Doyle framed the investment around regulated execution rather than another software layer that creates more work. Returning investors Y Combinator and Commerce Ventures bring early-stage and financial-services specialization, while new investor Fifth Wall has a lending-technology portfolio that includes companies working across real-estate finance.

From Missed Calls to Loan Operations

Kastle's origin sits inside a simple operating failure. Choudhary has said that, while working on Redfin's Mortgage Marketplace, he watched high-intent borrower leads reach lenders whose teams could not answer quickly enough. The consumer journey had moved online, but the operating layer behind it still depended on staffing, queues and systems built for a slower handoff.

Choudhary founded Kastle with Nitish Poddar, the company's CTO. Y Combinator lists Kastle as a San Francisco company founded in 2024 and part of its Summer 2024 batch. The accelerator's profile says Poddar previously led engineering work at Verkada, including motion-search and inference infrastructure, while Choudhary had worked on Redfin's mortgage marketplace and search products.

The founders began with AI voice and messaging agents for mortgage lenders, then widened the product toward loan operations. Kastle now presents separate agents for loan-officer assistance, collections, servicing, compliance and loss mitigation across voice, SMS, email, chat and APIs.

The Product Is the Work After the Conversation

A convincing voice agent can make a demo feel finished before the underlying operation has started. A real borrower interaction may require identity and loan context, payment processing, record updates, regulatory scripts, quality review, exception handling and a clean transfer to a person.

Kastle says its agents connect with existing contact-center, payment-processor and loan-management systems. The agents map to approved workflows, keep systems of record current and can route cases back to people. Its public site describes a compliance engine covering mortgage and consumer-lending requirements, along with simulation testing, interaction QA, observability and a SOC 2 Type II posture.

Those controls are central to the sales argument because lending obligations do not disappear when a model enters the workflow. The Consumer Financial Protection Bureau has made clear that creditors using complex algorithms for credit decisions still must provide specific and accurate reasons for adverse actions. Kastle's disclosed product scope is broader than credit decisioning, but the regulatory principle explains why banks will care as much about auditability and handoffs as conversational fluency.

Production Evidence Is Starting to Replace the Demo

Kastle says it is in production at 10 of the top 25 U.S. mortgage servicers. A July company-distributed announcement said eight, making the current website claim a later company-reported expansion rather than an independently measured market-share figure.

Named public deployments give the claim more texture. Carrington Mortgage Services disclosed autonomous servicing and collections agents plus tools that assist human contact-center teams. New American Funding later described a deployment spanning originations and servicing, with 24/7 borrower self-service and compliance controls.

Kastle's funding release says its agents have processed more than $1.8B in transactions, while the boilerplate on the same page says more than $2B. The lower figure is the defensible company-reported threshold. Kastle's site also advertises lower cost per call, reduced handle time and higher customer-satisfaction scores, but it does not publish an independent audit, customer-level methodology or error and escalation data for those metrics.

Why the Operating Boundary Matters

The Series A arrives as enterprise AI moves from generating material to executing work. In consumer lending, execution changes the risk profile. A system that writes a summary can be reviewed before it matters; a system that takes a payment, updates a borrower record or starts a collections workflow has already entered the operation.

That makes Kastle's boundary between automated capacity and human judgment commercially important. The company is not promising to remove people from every case. Its stated model uses agents for high-volume, repeatable work while employees handle situations that require expertise, empathy, relationships or exceptions.

The financing gives Kastle room to deepen that model across more institutions. Buyers will still need evidence around transaction accuracy, reversals, complaint handling, escalation rates, audit outcomes, security, retention and implementation economics. None of those details was disclosed with the round.

Kastle's opportunity is to make the old systems feel less like an obstacle without pretending they no longer matter. Each successful deployment will be judged in the moments where an agent completes the routine work, preserves the record and recognizes that the next decision belongs with a person.