At a Glance

Industry: Mortgage Technology (Secondary and Capital Markets)

Goal: Advance the AskLaura Guideline Assistant from proof of concept to a secure, scalable production service on AWS that delivers grounded mortgage guideline answers across LauraMac’s customer base.

OpsGuru Services: Agentic AI Enablement, Cloud Architecture, Production Readiness, Delivery Automation, Observability, and Knowledge Transfer

Tech Stack:

• AI Agent Platform: Amazon Bedrock AgentCore

• Knowledge Retrieval and Ingestion: Amazon OpenSearch Service, Amazon S3

• APIs and Integration: Amazon API Gateway, integration with LauraMac’s Java service

• Automation and Observability: Terraform, CI/CD pipelines, Amazon CloudWatch

Outcomes:

• Scalable service for multiple customers: LauraMac can expand the AskLaura Guideline Assistant across its customer base while maintaining secure separation of each client’s information and conversations.

• Strong answer quality: Evaluation on complex mortgage guideline questions demonstrated high accuracy, with responses grounded in approved source documents.

• Faster, responsive experiences: Streaming improvements help users begin receiving answers quickly. Reliable performance: Parallel testing demonstrated stable operation across evaluated production scenarios.

• Efficient operations: The AWS-based production design supports cost-effective growth.

Putting Complex Lending Guidelines Within Quick Reach

In the mortgage secondary and capital markets, loans are bought, sold, and packaged into securities, and every buyer has its own detailed rules about which loans it will accept. Those rules, known as underwriting guidelines, run long, change often, and exist in multiple versions at once. Finding the right answer for the right version can mean searching through many documents, and a wrong answer can mean a loan that does not meet an investor’s requirements.

LauraMac is a software-as-a-service company that provides an enterprise platform for the mortgage industry, with a focus on the secondary and capital markets. Its customers are mortgage businesses that need to make fast, accurate decisions with large volumes of policy information.

To enhance human decision-making, improve efficiency, and minimize errors through AI, LauraMac created AskLaura, its suite of AI-enabled agents embedded in the LauraMac platform. The suite includes the AskLaura Guideline Assistant, an agent that answers questions about underwriting guidelines by retrieving relevant material from a knowledge base of the customer’s documents and generating answers grounded in those sources.

A proof of concept, built with OpsGuru on an initial AWS foundation, demonstrated the agent’s value and user demand.

The proof of concept had been designed around a single client. To offer the assistant across its customer base, LauraMac needed a production service that could serve many customers at once, each with its own documents and conversations kept private. LauraMac selected OpsGuru, its partner on the earlier proof-of-concept work and the team that introduced it through its AWS account, to take the assistant the rest of the way to production.

Scaling a Successful Proof of Concept to Production

Scaling the AskLaura Guideline Assistant meant preserving the security, accuracy, and usability expected of an enterprise mortgage platform.

• Building on a Successful Proof of Concept: Early work demonstrated that the AskLaura Guideline Assistant could make complex underwriting information easier to use. The next step was preparing the service for broader production adoption.

• Secure, Customer-Specific Experiences: Each customer needed answers grounded in its authorized guideline content, with secure separation of information and conversations.

• Answers That Hold Up Under Scrutiny: Underwriting guidance depends on the applicable policy and context. The AskLaura Guideline Assistant needed to remain grounded in approved sources and clearly acknowledge when information was unavailable.

• Production-Grade Operations: The service required repeatable delivery, monitoring, and operational visibility to support dependable use and continued growth.

The Secure, Scalable AWS Agentic AI Foundation for the AskLaura Guideline Assistant

Phase 1: Discovery and Production Design

LauraMac and OpsGuru worked through a structured discovery and delivery process to align business, security, and operational requirements. Together, the teams established an AWS-based production design that would support the AskLaura Guideline Assistant’s growth while preserving LauraMac’s product ownership and customer commitments.

Phase 2: Production Engineering and Integration

• Scalable AI Service: LauraMac and OpsGuru implemented a production architecture that enables the AskLaura Guideline Assistant to serve multiple customers while maintaining secure separation of client information and experiences.

• Grounded Knowledge Experience: The solution uses a retrieval-augmented generation (RAG) approach to keep the AskLaura Guideline Assistant’s answers connected to authorized mortgage guideline content and supports controlled updates as source materials change.

• Application Integration: Secure service interfaces enable LauraMac’s platform to deliver the AskLaura Guideline Assistant within its existing customer experience. OpsGuru worked in collaboration with LauraMac's engineering team on the production integration.

• Automated Delivery and Observability: Repeatable deployment, monitoring, and alerting capabilities help LauraMac operate the service consistently and identify issues early.

Phase 3: Validation, Production Readiness, and Handover

LauraMac and OpsGuru validated the service through acceptance, integration, and performance testing, adapting the implementation as requirements evolved. OpsGuru supported production readiness and handover. Documentation and knowledge-transfer sessions prepared LauraMac’s engineering and operations teams to manage, monitor, and extend the AskLaura Guideline Assistant.

Bringing AskLaura to production required translating an innovative product vision into a secure, scalable service. AWS and OpsGuru partnered closely with our team to design and deliver the architecture, adapting at every stage, while we retained ownership of the product and customer experience. — Amit Aggarwal, Chief Product and Technology Officer, LauraMac

The Outcome: Trusted, Responsive AI in Production

The AskLaura Guideline Assistant is live in production, helping LauraMac advance an AI-enabled customer experience for navigating complex mortgage guideline information.

Technical Outcomes: Grounded, Responsive, and Reliable

• Accurate, grounded answers: Evaluation on complex mortgage guideline questions demonstrated high accuracy, with responses grounded in approved source documents and appropriate handling when relevant information was not available.

• Responsive user experience: Streaming improvements allow users to begin receiving answers quickly, creating a more natural and productive interaction.

• Reliable operation: Parallel testing demonstrated stable performance across evaluated production scenarios.

• Efficient foundation: The AWS-based production design supports cost-conscious operation as usage grows.

Business Outcomes: A LauraMac-Owned Service Ready to Grow

• Scalable customer value: LauraMac can extend the AskLaura Guideline Assistant across its customer base while maintaining secure separation of client data and information.

• Production-Ready and owned by LauraMac: The AskLaura Guideline Assistant is live in production, supported by automated delivery, operational monitoring, documentation, and knowledge transfer that enable LauraMac’s teams to manage and evolve the service independently.

• A collaborative delivery partner: OpsGuru worked closely with LauraMac and AWS, adapting the production approach as requirements evolved while supporting LauraMac’s product vision.

• A foundation for continued innovation: The production foundation gives LauraMac flexibility to expand the AskLaura Guideline Assistant’s capabilities and customer value over time.