In the realm of investing, stakeholders assess key factors to evaluate a company’s potential for growth, risk and sustainability. The Environmental, Social and Government (ESG) criteria is a crucial component of their assessments and offers a comprehensive perspective on a company's practices and their repercussions. However, ESG audits involve a labour-intensive process of gathering and reviewing various sources and documents, followed by assessing and processing the accumulated information, before generating a high-quality report.

To streamline this process, researchers from A*STAR Institute of High Performance Computing (A*STAR IHPC) developed EnterpriseFind, an enterprise tool for automating operations, human resource and audit processes. By leveraging Generative AI (Gen AI) and Retrieval Augmented Generation (RAG) capabilities, EnterpriseFind produces comprehensive reviews for enterprises, such as ESG reports, optimising time and cost resources.

Features

• Effortless Data Exploration: Ingests and processes various reports and internal documents, ensuring seamless retrieval of required information. making it seamless to retrieve required information.
• Powerful Semantic Search: Extends beyond simple keyword matching and provides relevant, insightful answers.
• AI-powered Response Generation: Generates clear and concise responses tailored to specific queries by using a Language Model.
• Customisable Framework: Integrates seamlessly with user’s existing data infrastructure, ensuring a smooth and customised user experience.
• Transparent through Extractive Explanations: Identifies relevant information in the document and provides page numbers, facilitating easy reference and fact checking

• Document Processing & Embeddings: These AI models convert textual data from uploaded documents into embeddings, where each sentence becomes a unique point in a high-dimensional space, while similar sentences are positioned close together.
• Semantic Matching & Vector Search: By transforming user queries into embeddings as well, EnterpriseFind efficiently searches the Vector Database (which contains document embeddings) to find the closest matches. This allows the system to identify relevant content based on semantic meaning rather than just exact keywords.
• Response Generation & Large Language Models (LLMs): LLMs are trained on extensive text data, allowing them to understand complex relationships between words and generate human-quality text. The framework provides contextual information to the LLM from relevant documents and prompts it to generate a response that coheres with the retrieved information.

• HR internal document processing
• Healthcare document retrieval
• Finance document retrieval