Kastle

What do they actually do

Kastle builds AI agents that handle borrower-facing servicing tasks for mortgage lenders and servicers over voice calls, SMS, chat and email. The system plugs into a servicer’s telephony and loan‑servicing platforms so the AI can read loan details, verify identity, answer payoff/escrow questions, take payments, post notes, make allowed account changes, and escalate to a human when needed (kastle.ai; YC).

The product runs as a controlled workflow engine that maps lender SOPs into “action blocks,” ensuring the agent follows approved scripts and only performs specific actions. All interactions are recorded and scored against a lender’s QA/compliance rules for auditability. Kastle emphasizes prebuilt regulatory guardrails, simulation before go‑live, and audit tooling to reduce production risk (kastle.ai; Forbes; Tidalwave).

Evidence to date is early but visible: the company is running pilots and announced a partner relationship with Haven, and it won LendingTree’s 2024 Innovation Challenge for voice AI demos. Kastle was founded in 2024 and is a YC S24 company (FintechFutures; PR Newswire; YC).

Who are their target customer(s)

• Mortgage servicers at banks and non‑bank servicers: High recurring contact‑center costs and strict compliance demands; inconsistent human agent behavior and incomplete logs increase audit risk.
• Sub‑servicers and white‑label servicing platforms: Need standardized workflows across many lender clients; customizing integrations and SOPs at scale while keeping every client’s calls auditable is costly and complex.
• Collections and loss‑mitigation teams: Sensitive outbound calls and negotiations carry legal/financial risk; teams must follow strict scripts, document thoroughly, and handle volume spikes without proportional headcount increases.
• Contact‑center / CX leaders at lenders: Hiring, training, and QA for large agent pools is expensive; high turnover and limited ability to score more than a sample of interactions hinder consistency and compliance.
• Loan‑origination and sales teams (near‑term): Slow or missed lead follow‑up and uneven qualification waste loan officer time; manual triage reduces conversion and delays borrower response.

How would they acquire their first 10, 50, and 100 customers

• First 10: Run 8–12 week paid pilots with early‑adopter servicers/sub‑servicers, integrating to their telephony and servicing systems to prove compliance, QA scoring, and safe account actions in production‑like simulations. Leverage demos, the LendingTree award, and announced partner pilots to secure these initial deals (kastle.ai; PR Newswire; FintechFutures).
• First 50: Expand via partnerships with servicing platforms/resellers and ship prebuilt “action block” templates and integration packs to shorten onboarding and reduce risk. Use early customers’ audit trails and QA results as references to accelerate procurement at peer servicers (kastle.ai; FintechFutures; Forbes).
• First 100: Stand up an enterprise sales motion for large servicers while growing a channel of contact‑center vendors and sub‑servicers who can white‑label Kastle. Use dedicated success/implementation squads and standardized compliance playbooks for multi‑site rollouts; sell verticalized bundles (collections, loss‑mit, origination triage) for clear ROI (kastle.ai; Tidalwave; Forbes).

What is the rough total addressable market

Top-down context:

There are roughly 54 million first‑lien residential mortgages in the U.S., inferred from OCC data showing 10.8 million loans at seven major banks representing 20% of the national total (OCC Mortgage Metrics Q2 2025). MBA data indicates annual servicing cost benchmarks of about $176 per performing loan and $1,857 per non‑performing loan in 2023, with call center included in base servicing costs (MBA Newslink; Progress in Lending summarizing MBA).

Bottom-up calculation:

Using ~54M loans and OCC’s 97.5% current rate: ~52.65M performing × $176 ≈ $9.3B plus ~1.35M non‑performing × $1,857 ≈ $2.5B → total servicing cost ≈ $11.8B/year. If 20% of servicing spend is borrower contact‑center/QA/communications addressable by Kastle, U.S. SAM ≈ $2.4B/year; expansion into origination front‑end would increase this figure (OCC; MBA/Progress in Lending).

Assumptions:

• OCC’s 10.8M loans equal ~20% of U.S. first‑lien mortgages, implying ~54M total; 97.5% current vs. 2.5% delinquent (OCC Q2 2025).
• MBA 2023 annual cost per loan: $176 performing, $1,857 non‑performing; call center is part of base servicing costs (MBA Newslink; Progress in Lending).
• 20% of servicing costs are attributable to borrower contact‑center/QA/communications functions that Kastle can replace or automate; U.S. only.

Who are some of their notable competitors

• Replicant: Voice and chat autonomous contact‑center platform used across verticals (including mortgage). Overlaps on automating borrower calls; Kastle differentiates by tighter servicing‑system integrations and action‑centric workflows (use case).
• Observe.AI: AI agents plus auto‑QA and conversation intelligence for regulated industries. Strong in compliance scoring and agent assist; often complements human teams as much as full automation (AI agents).
• Cresta: Real‑time coaching, agent assist, and automated conversational agents for financial services. Emphasizes human agent guidance and operations oversight alongside automation (AI agent).
• TrueAccord: Digital collections platform automating outreach across email/SMS/web for compliant debt recovery. Focuses on digital journeys vs. telephony‑integrated, action‑posting servicing agents.
• Interactions (IVA): Enterprise IVA vendor with bespoke deployments in financial services and collections. Incumbent in large custom conversational AI projects; Kastle positions as mortgage‑specific with built‑in SOP/action blocks and audit tooling.