Discussing the next frontier of nonprofit innovation — an interview with Scott Graham and Sri Elaprolu.

At FINCA we think cutting-edge tools like AI can be a powerful force for social good. FINCA International’s VP of Research and Data Science Scott Graham sat down with the Director of the AWS Generative AI Innovation Center Sri Elaprolu to talk about how FINCA’s new Automated Research Assistant (AuRA), developed in collaboration with AWS, is harnessing AI to streamline research and generate actionable insights for creating pathways out of poverty.

In working with AWS, FINCA has embraced AI as a transformative force on our path to eradicating global poverty. Though we are still at an early stage of our journey with this groundbreaking technology, we are already seeing its potential to shape the future of financial inclusion and social impact solutions.

FINCA is developing an AI-powered innovative platform to enhance our research and drastically accelerate how we identify evidence-backed poverty solutions. Scott and Sri unpack how AuRA works in their interview (embedded below). In summary, AuRA operates in three phases:

1. Amplification

The first phase begins when a researcher poses a general query about a specific topic to a foundational model (in our case, we use Amazon Bedrock). FINCA’s new platform AuRA then generates an outline of themes and subtopics relevant to the issue — what’s called a topic model — and automatically searches them across multiple databases using Apache Airflow.

2. Document Retrieval

In the second phase, AuRA uses process automation to compile a robust document library that indexes the pertinent research materials to each section of the topic model.

3. Curation

In the final and most critical stage, human researchers refine and validate the AI-generated topic model and sourced documents to ensure the resulting research framework is both accurate and trustworthy. AuRA then reprocesses the curated insights using retrieval-augmented generation to refine an impact pathway that is fully rooted in the context of our verified information sources.

This article first appeared on finca.org.