KKastle

Kastle builds AI agents that pick up mortgage borrowers’ calls, take payments, open servicing tasks and work early collections inside the systems lenders already run. In 2026 it signed Carrington Mortgage Services, with more than $200 billion in unpaid balances, and New American Funding, then raised a $24 million Series A led by Insight Partners. Kastle says its agents have processed more than $1.8 billion in transactions.

It fits servicers, banks and large lenders with heavy borrower call volume and systems such as ICE MSP. Pricing is custom, and the public customer list is made up of large institutions, so small brokers may want a lighter tool. We rate it 7.8/10 on our scale, with user sentiment at a neutral score because review sites do not yet rate it.

Kastle by the Numbers

Kastle is two years old, so most public figures come from trade coverage of its customer deals and funding round.

The two customer deals are the evidence that matters. Carrington and New American Funding are both large, regulated mortgage companies, and both picked Kastle for borrower-facing work. Carrington’s EVP of loan servicing said it needed proven performance at scale; NAF’s president said Kastle let it tailor agents and test them before launch.

The investor list adds a signal. Insight Partners led the round, Y Combinator and Commerce Ventures returned, and Fifth Wall, a real estate technology investor, joined. An Insight Partners executive said Kastle deploys agents that can pass the bar on regulatory rigor, which is the standard large servicers apply.

The $1.8 billion transaction figure shows money already moving through the agents, which a servicer’s risk team will ask about. What is not public yet are call containment rates and borrower satisfaction scores, so ask for both in a pilot.

How Kastle Agents Handle a Servicing Call

Kastle started with an agent operating system for mortgage servicers, built around voice conversations, and has since added origination support. The agents can run on their own or sit next to human staff as assistants.

Inside the Servicing System

The April 2026 integration with ICE’s MSP, the servicing system with the largest market share, lets agents retrieve loan data, process payments, create servicing tasks, log activity and run collections-related workflows. Kastle also says its agents work across legacy core systems, keeping systems of record updated so lenders do not have to replace them.

Safe Execution Procedures

Regulated, multistep transactions run under what Kastle calls Safe Execution Procedures. A borrower who asks to change a payment date, for example, goes through identity checks and the lender’s rules in a fixed order, and the agent records each step. NAF’s president highlighted the ability to tailor agents and test them before launch.

The Compliance Engine

Kastle’s compliance engine covers RESPA, TILA, TCPA and FDCPA, plus the ban on unfair, deceptive or abusive acts. Every interaction is scored against the lender’s quality standards, which gives compliance teams a record of each call.

Collections and Quality Control

At Carrington, the scope includes autonomous customer service and collections agents, assistants for employees and quality control coverage. Kastle’s agents give real-time responses based on data analysis and past interactions, so a borrower calling back does not have to repeat the history.

Origination Calls

At New American Funding, Kastle agents cover originations as well as servicing. NAF’s stated goal is to handle more calls without its operations teams absorbing all of the added volume, while keeping conversations that need a human touch with staff.

Assistant Mode for Staff

Kastle’s agents do not have to face borrowers. In assistant mode they work beside human staff, pulling data and drafting answers while an employee stays on the line. Carrington’s deployment includes employee assistance alongside the autonomous agents, which gives a servicer a gentler first step: staff get help on live calls before any call is handed fully to an agent.

Testing Before Launch

NAF’s president said the lender chose Kastle partly for the ability to tailor agents and test them before launch. In practice, that means running agents against sample calls for each procedure, checking the scores, and only then routing live borrowers. For a regulated servicer, this pre-launch stage is where compliance teams sign off.

A Borrower Call From Start to Finish

A borrower calls on a Sunday evening to make a late payment and ask about escrow. The Kastle agent verifies identity, pulls the loan from MSP, takes the payment and logs it, then explains the escrow line using the loan data. The borrower mentions a recent job loss. The agent follows the lender’s procedure for hardship mentions, creates a servicing task for the loss mitigation team and tells the borrower when to expect contact. On Monday, a specialist opens the task with the call summary and score already attached.

Work Kastle Leaves to Servicing Staff

Judgment calls, relationships and empathy stay with people, according to Kastle’s own description of hybrid teams. Hardship decisions, complex escalations and anything outside the defined procedures go to staff. The lender writes and approves the procedures.

Kastle Pricing Through Custom Contracts

Kastle does not publish prices. Engagements are scoped with each institution.

Modeling the Case for a Servicer

Build the case from your own call data. A servicer with 100,000 inbound calls a month and an average handle time of 6 minutes spends 600,000 agent minutes, or 10,000 hours, on the phone. If agents fully handled 30 percent of those calls, that frees 3,000 hours a month. At a loaded cost of $35 an hour, that is $105,000 a month in staff time. The 30 percent is an assumption for illustration; ask Kastle for containment data from a pilot on your call mix.

A smaller lender with 10,000 calls a month at the same handle time spends 1,000 hours. The same 30 percent would free 300 hours, worth $10,500 a month at $35 an hour. Below that scale, a custom enterprise contract may be harder to justify.

Remember the cost side of the ledger. Agent minutes, telephony, procedure writing and compliance review time all belong in the model, along with the internal project team needed for testing. A pilot on a few call types keeps those costs bounded while the containment rate is measured.

Factor in integration work. ICE MSP has a direct connection; other systems may need more setup.

Kastle Compared With Sei AI, Salient and Structurely

A useful way to choose is to map where the call volume sits. If most calls arrive after closing, about payments, escrow and payoffs, a servicing-first vendor such as Kastle matches the need. If the pressure is on new leads and applications, a sales-focused agent fits better, and if documents and QC are the bottleneck, a vendor with document agents earns a look.

Kastle and Sei both target mortgage operations. Kastle’s public deals center on servicing at large scale; Sei covers origination, underwriting documents and QC as well. Salient serves consumer and auto lending call centers. Structurely works earlier in the funnel, following up with leads. Our mortgage and lending agents page and banking agents page list more.

Lenders Who Fit Kastle

Mortgage servicers and subservicers with large call centers fit best, especially those on ICE MSP. Payments, escrow questions and early collections are high volume and rule-bound, which suits agents.

Lenders that both originate and service loans can use one vendor across both, as NAF does.

Servicers growing through portfolio acquisitions also fit. Carrington added roughly 800,000 loans when it bought Valon’s servicing portfolio, and agents can absorb call spikes from a transfer of servicing without a hiring wave.

Banks and credit unions with consumer lending books and legacy cores also fit, since Kastle works across existing systems.

When Another Option May Suit Better

Lenders that are not ready to write detailed procedures for each call type may also want to wait, since Kastle’s model depends on them.

Small brokers and loan officer teams focused on new leads may get more from a lead follow-up tool such as Structurely, which publishes reported pricing.

Lenders that want document processing, underwriting income calculation and QC from the same vendor may look at Sei AI. Auto lenders may prefer Salient’s focus.

Pilot Metrics, Procedures and Security Questions

Ask for containment, transfer and complaint rates from a pilot on your own call types before scaling. Public reports name customers but not outcome metrics.

Procedures are only as good as the lender’s own. Spend time writing and testing Safe Execution Procedures for each call type, and involve compliance from the start.

Outbound calling brings TCPA consent and call-time rules into play, and collections calls bring FDCPA limits. Kastle’s compliance engine covers both laws, but the lender still owns consent records and calling windows, so confirm how those flow into the agent’s procedures.

Borrowers’ comfort with AI is another factor. NAF framed its deployment as keeping human conversations where they matter, and lenders should give borrowers a clear path to a person on every call.

Kastle could publish its security certifications and data handling terms on its site. Until then, request them early, since procurement at regulated lenders will require them.

Kastle’s 2026 Deals and Series A

In April 2026, Kastle announced its direct integration with ICE’s MSP. On July 15, 2026, National Mortgage News reported that Carrington Mortgage Services would deploy Kastle agents in its servicing operations. On August 20, New American Funding partnered with Kastle for voice agents across originations and servicing. On September 18, Kastle raised a $24 million Series A led by Insight Partners.

Our Take on Kastle for Mortgage Servicing

Kastle earns 7.8/10 on our scale. Autonomy is high because agents finish payment and servicing tasks inside MSP under defined procedures. Trust scores well for its built-in compliance engine and call scoring. Pricing value and setup sit lower because pricing is custom and deployments are enterprise projects.

The score has room to rise. Published certifications, outcome metrics from live deployments and a public review profile would lift trust and sentiment, and a packaged offer for mid-size lenders would lift pricing value and setup.

For a servicer or large lender with high call volume, Kastle is a credible option with named enterprise customers. Run a pilot on two or three call types, measure containment and complaints, and expand from there.

Servicer Questions About Kastle

How Much Does Kastle Cost?

Kastle does not publish pricing. Engagements are scoped and priced with each bank, servicer or lender, and Kastle pitches the agents as a way to add capacity without adding headcount, so expect the quote to be framed against staffing costs.

Does Kastle Integrate With ICE MSP?

Yes. Since April 2026, Kastle agents connect directly to ICE’s MSP to retrieve loan data, process payments, create servicing tasks, log activity and run collections workflows.

Which Regulations Does Kastle Cover?

Kastle’s compliance engine covers RESPA, TILA, TCPA and FDCPA, plus rules against unfair, deceptive or abusive practices. Every interaction is scored against the lender’s standards.

Who Uses Kastle?

Public customers include Carrington Mortgage Services and New American Funding. Kastle also says it works with major enterprises and banks, and it plans to accelerate deployments with banks and other financial institutions across North America after its Series A.

Can Kastle Handle Hardship and Collections Calls?

Kastle runs collections-related workflows, and Carrington’s deployment includes collections agents. Hardship decisions and complex cases follow the lender’s procedures, which typically route them to staff.