Enterprise-Grade Lending AI Agents for a Consumer Loan Provider
Challenge
The customer is a US-based consumer lending company that provides personal, credit-starter and signature loans to individuals, including those from underserved communities. It already had a robust proprietary loan management system automating core loan processing and servicing, but with recent advances in large language models and agentic AI it saw an opportunity to further improve operational efficiency and profitability. The customer needed a partner to validate the feasibility of agentic AI for its business, plan agentic solutions and deploy them safely within its regulated lending workflows, with strong governance, data security and compliance. Having worked with INNERLUXES's lending software engineering team on its core systems for several years, it turned to a trusted partner that already knew its environment and had a long track record in lending AI engineering.
The customer is a US-based consumer lending company that provides personal, credit-starter and signature loans to individuals, including those from underserved communities. It already had a robust proprietary loan management system automating core loan processing and servicing, but with recent advances in large language models and agentic AI it saw an opportunity to further improve operational efficiency and profitability. The customer needed a partner to validate the feasibility of agentic AI for its business, plan agentic solutions and deploy them safely within its regulated lending workflows, with strong governance, data security and compliance. Having worked with INNERLUXES's lending software engineering team on its core systems for several years, it turned to a trusted partner that already knew its environment and had a long track record in lending AI engineering.
Solution
INNERLUXES's consultants ran focused workshops to understand operational needs and constraints, then scoped high-impact agentic use cases, each with a solution concept, a software requirements specification and tailored performance, quality and user-acceptance metrics. The customer prioritized three task-specific agents: a loan application verification agent that voice-calls applicants who passed credit-bureau checks to verify identity and confirm borrowing intent; a debt collection agent that guides debtors through repayment options while adhering to FDCPA regulations; and an incoming call handling agent that triages inbound calls and routes them to the right specialist. A feasibility study showed that building custom agents on cloud AI services would be cheaper and more controllable than customizing packaged tools, so the customer opted for a custom, incremental build. For the first agent, architects used LiveKit Cloud and the LiveKit Agent SDK on Azure Container Apps, designing for cost-effective scaling, sub-250ms low latency, interoperability, deterministic agent governance with full observability, zero-trust security, and region-specific deployments for TCPA and GLBA compliance. Engineers built and integrated the agent with the customer's website and admin panel, and QA validated it through functional, integration, end-to-end, edge-case and performance testing.
INNERLUXES's consultants ran focused workshops to understand operational needs and constraints, then scoped high-impact agentic use cases, each with a solution concept, a software requirements specification and tailored performance, quality and user-acceptance metrics. The customer prioritized three task-specific agents: a loan application verification agent that voice-calls applicants who passed credit-bureau checks to verify identity and confirm borrowing intent; a debt collection agent that guides debtors through repayment options while adhering to FDCPA regulations; and an incoming call handling agent that triages inbound calls and routes them to the right specialist. A feasibility study showed that building custom agents on cloud AI services would be cheaper and more controllable than customizing packaged tools, so the customer opted for a custom, incremental build. For the first agent, architects used LiveKit Cloud and the LiveKit Agent SDK on Azure Container Apps, designing for cost-effective scaling, sub-250ms low latency, interoperability, deterministic agent governance with full observability, zero-trust security, and region-specific deployments for TCPA and GLBA compliance. Engineers built and integrated the agent with the customer's website and admin panel, and QA validated it through functional, integration, end-to-end, edge-case and performance testing.
Results
INNERLUXES completed full production deployment of the loan application verification agent across all of the customer's branches after successful user acceptance testing, where it accurately handled verification, maintained stable conversational flows and moved data smoothly across connected systems. The engagement launched an agentic AI solution tailored to the company's loan-processing requirements and compliance rules, reduced workload for call-center and lending teams by offloading repetitive verification and communication tasks, and enhanced lending risk control through automated applicant validation at early stages. The debt collection agent is estimated to increase debt recovery rates by up to 20% while lowering collection costs. A modular architecture, pragmatic tech stack and fast iterative delivery enabled a quick pilot rollout and early business value, with secure, compliant customer interactions, full auditability of agent actions, and a scalable foundation for further AI adoption. INNERLUXES has since begun the technical design of the debt collection agent, applying the same architectural principles.
INNERLUXES completed full production deployment of the loan application verification agent across all of the customer's branches after successful user acceptance testing, where it accurately handled verification, maintained stable conversational flows and moved data smoothly across connected systems. The engagement launched an agentic AI solution tailored to the company's loan-processing requirements and compliance rules, reduced workload for call-center and lending teams by offloading repetitive verification and communication tasks, and enhanced lending risk control through automated applicant validation at early stages. The debt collection agent is estimated to increase debt recovery rates by up to 20% while lowering collection costs. A modular architecture, pragmatic tech stack and fast iterative delivery enabled a quick pilot rollout and early business value, with secure, compliant customer interactions, full auditability of agent actions, and a scalable foundation for further AI adoption. INNERLUXES has since begun the technical design of the debt collection agent, applying the same architectural principles.