SmartSearch Chatbot for Document Retrieval

50+

Monthly Staff Hours Saved

0

Downtime During Failure Events

Background

FINCA International is a global nonprofit focused on financial inclusion and poverty alleviation. Founded in 1984, it provides microloans and supports social enterprises in over 15 countries, helping underserved communities access financial services, energy, education, and healthcare. FINCA is known for empowering low-income entrepreneurs—especially women—through responsible lending and innovative, tech-driven solutions.

Challenge

FINCA employees faced significant challenges in quickly accessing critical internal information, often spending several hours manually searching across various systems and data formats such as text, JSON, CSV, and mixed media. This inefficiency led to delays in decision-making and diverted attention from their core mission of supporting financial inclusion in underserved communities. To address this, a document retrieval chatbot was developed to allow natural language queries across internal document repositories, improving accessibility and streamlining operations. The solution integrated secure authentication through Microsoft Entra and leveraged tools like AWS Lambda, Bedrock Knowledge Bases, Langchain, Langraph, and OpenSearch with Cohere embeddings, along with Terraform-based deployment. While the chatbot had limitations in summarizing tabular data, it successfully delivered relevant responses, significantly reducing search time and improving employee productivity.

Approach

To address FINCA’s need for faster, more reliable access to internal knowledge, we designed and deployed a fully serverless, scalable document retrieval chatbot using AWS-native services. The solution enables employees to ask natural language questions and receive relevant document insights in real time, replacing hours of manual searching with instant responses. Key components include AWS Lambda for orchestration, Amazon Bedrock Knowledge Bases for querying Claude 3.5 models, OpenSearch Serverless with Cohere embeddings for semantic search, and DynamoDB for chat history and feedback. Microsoft Entra ID was integrated for secure authentication and access control, aligning with FINCA’s enterprise identity standards. The architecture was built with high availability, fault tolerance, and scalability in mind. AWS Amplify and CloudFront serve the frontend with global edge redundancy, while API Gateway and Lambda handle backend logic with seamless auto-scaling. All components were deployed across multiple Availability Zones to eliminate single points of failure. Terraform was used to automate deployment, ensuring rapid, repeatable environment provisioning. This approach provided FINCA with a low-maintenance, high-impact solution that meets both their operational and security requirements, while laying the groundwork for future enhancements such as tabular data summarization.

Key results

Significantly Improved Access to Internal Knowledge: The chatbot transformed how FINCA employees locate and use institutional knowledge. Previously, retrieving relevant information from internal documents could take hours of manual searching through various systems and file formats. With the new AI-powered chatbot, users can now ask natural language questions and receive accurate, context-rich answers in real time. This shift has accelerated decision-making across teams, reduced administrative burden, and allowed staff to focus more on their core mission—advancing financial inclusion in underserved communities.

Secure and Seamless User Experience Aligned with Enterprise Standards: The solution was designed with strict security requirements in mind, integrating directly with Microsoft Entra ID (formerly Azure AD) for enterprise-grade identity management. This allowed FINCA to enforce single sign-on (SSO), conditional access policies, and multi-factor authentication without disrupting user workflows. Employees can securely access the chatbot via a web-based frontend hosted on AWS Amplify, with all API interactions authenticated via JWT tokens issued by Entra. This approach ensured that access control and data privacy met FINCA’s internal governance standards while keeping the user experience simple and intuitive.

Resilient, Scalable Architecture That Requires Minimal Maintenance: The chatbot was architected as a fully serverless solution, using AWS-native services such as Lambda, API Gateway, Bedrock Knowledge Bases, OpenSearch Serverless, and DynamoDB—all deployed across multiple Availability Zones for high availability. The use of Terraform for infrastructure as code enabled fast, repeatable deployments and simplified environment management. This design not only ensured fault tolerance and elasticity to handle usage spikes, but also minimized operational overhead, making the solution both cost-effective and sustainable for long-term use.

Benefits

'This project demonstrates how a thoughtfully designed, AI-powered solution can create meaningful impact for mission-driven organizations like FINCA. By drastically reducing the time employees spend searching for information, the chatbot allows teams to focus more on their core work—advancing financial inclusion in underserved communities. With a secure, scalable, and low-maintenance architecture, the solution not only met FINCA’s technical and operational goals but also laid a strong foundation for future innovation. It’s a clear example of how modern cloud technologies can directly support social impact at scale.

Technology partners

AWS

Amazon Web Services (AWS) is a global leader in cloud-computing services that offers over 200 fully featured services from data centers worldwide. AWS provides services to organizations of all sizes to help lower costs, become more agile, and innovate faster.

Advanced Tier Partner

with a Data & Analytics Competency

30+

General Certifications

Speciality and Professional Certifications