Base39 cuts loan analysis costs by 96% with MongoDB

THEIR CHALLENGE

Credit analysis complexities and lost opportunities at Base39

A financial technology provider based in São Paulo, Brazil, Base39’s core offerings are centered on providing advanced services for credit and risk analysis. Base39 was established to disrupt credit analysis by harnessing the power of artificial intelligence with MongoDB Atlas Vector Search and Amazon Bedrock.

Base39 realized that existing credit analysis methods produce financial decisions that are insufficiently comprehensive, primarily due to data scarcity. Essential data—such as income verification, employment history, and credit scores—represent a fraction of the information required. Furthermore, gathering and selecting these data points is heavily reliant on subjective processes. This complex and highly manual process can take up to 10 days and can yield inaccurate and incomplete results.

Furthermore, these time- and labor-intensive tasks require skilled subject matter experts with intricate understanding of using machine learning (ML) models. Some of Base39’s largest clients are accustomed to employing data scientists to work with traditional ML models—yet they still struggled to capitalize on loan approvals. Adapting these models to specific rules can take over a month due to the extensive knowledge required to analyze any new data source.

With a mission to deliver a service that enables its customers to analyze and assess up to 10 loan applications in a minute, Base39 sought a solution that would accelerate the review process by removing these technical and analytical constraints.

OUR SOLUTION

Transforming credit analysis with agentic AI and MongoDB Atlas

To modernize the loan assessment process, Base39 selected MongoDB Atlas on Amazon Web Services (AWS) as their foundational platform. They integrated agentic AI, orchestrated by LangGraph, to interact with machine learning models and large language models (LLMs) like Claude 3.7 Sonnet and Claude 3.5 Haiku via Amazon Bedrock.

MongoDB Atlas’ native vector search capabilities allows Base39 to address challenges in data storage and retrieval effectively. Instead of manual processes involving spreadsheets and unstructured documents, the solution leverages LLMs to provide interactive, AI-driven tools for loan risk assessors. It aggregates relevant data from sources like CRM systems, transactional records, and public databases, and offers recommendations on cleaning, aggregating, and transforming data for risk evaluation.

Predictive machine learning algorithms handle basic loan scoring, while generative AI (gen AI) enhances data processing with historical analysis, dynamic model updates, and "what-if" evaluations. This eliminates inefficiencies, with the LLM acting as a smart assistant to guide risk analysts in selecting key data sources and fields for specific scenarios.

A defining feature of Base39’s solution is its agentic AI approach, enabling autonomy in data analysis and decision-making. Using a Chain-of-Thoughts (CoT) methodology, the system reasons, perceives, and acts based on loan-specific data.

MongoDB Atlas serves as the backbone of Base39’s feature store, powering both structured and unstructured data analysis with MongoDB Atlas Vector Search. This technology accelerates feature retrieval, dynamically updates machine learning models, and provides real-time insights for credit policy adjustments. By enriching LLM recommendations with domain-specific context, Base39 ensures risk assessments are accurate, reliable, and efficient.

MongoDB's flexible schema supports diverse data types, making model updates seamless and adaptable. Combined with Atlas's serverless architecture, Base39 reduces operational overhead, freeing development teams to focus on innovation while ensuring credit models remain responsive.

OUTCOME

Increased productivity, streamlined processes, and reduced costs

The strategic partnership with MongoDB and AWS enabled Base39 to deploy a robust AI-driven solution in two weeks. By shifting from manual processes to an AI-driven solution, Base39 cut loan analysis costs by 96% and reduced infrastructure expenses by 84%. Additionally, MongoDB Atlas's serverless architecture reduces operational burdens on Base39’s development teams. Since implementing the new solution, developers have been able to focus on innovation instead of infrastructure management.

By setting a new standard for efficiency and innovation in loan management, Base39 delivers exceptional value to its customers, ensuring no loan opportunity is missed. Their strategic shift positions Base39 as leaders in adopting AI to disrupt the financial services industry.

To learn how you can build modern, cloud-based solutions, visit MongoDB Atlas on AWS.