SAS launches retrieval agent manager for enterprise unstructured data processing

Analytics company releases no-code AI solution to process unstructured enterprise data using retrieval augmented generation framework. The platform aims to help organizations extract information from documents, manuals, and reports for business decision support across multiple industries.

SAS has released Retrieval Agent Manager (RAM), a platform designed to process unstructured enterprise data and generate responses for business applications. The solution addresses the challenge of managing unstructured data formats such as text and images, which represent over 80% of enterprise data and grow 50-60% annually according to industry estimates.

RAM operates on a retrieval augmented generation framework as a no-code solution that processes unstructured documents and configures systems for document interaction via APIs or chatbots. The platform supports integration with generative AI services including large language models and vector databases, and includes an agentic AI layer for workflow automation using enterprise data.

Kathy Lange, Research Director for AI and Automation at IDC, stated that the platform converts unstructured information into enterprise knowledge through generative and agentic AI, providing an interface for process development without requiring system overhauls.

The solution ingests and processes unstructured documents, evaluates configurations for document interaction, and generates responses based on enterprise content. RAM maintains separation between enterprise data and large language models rather than using company data for model training or fine-tuning, creating a knowledge service that combines corporate data with language models to generate responses.

Financial services applications include fraud pattern detection and regulatory compliance support for anti-money laundering and know-your-customer processes. Risk management teams can retrieve stress test models, policy documents, and loss histories for exposure assessment and capital planning under regulatory frameworks.

Insurance applications include policy language retrieval, first notice of loss documentation access, and prior claim file review to support claims resolution and regulatory compliance. Public sector implementations enable contact center agents to access information from service tickets, policy manuals, and case archives for citizen service responses.

Healthcare applications allow clinicians to access patient notes, clinical protocols, and research documentation while maintaining HIPAA compliance requirements. The platform synthesizes information from multiple medical sources to support clinical decision-making.

Manufacturing implementations focus on predictive maintenance applications. The platform processes maintenance manuals, inspection reports, vendor records, service bulletins, and documentation to support equipment issue diagnosis and response planning. RAM retrieves repair and maintenance information to generate work orders for engineering and technical staff, complementing existing machine learning-based predictive maintenance systems.

The agentic AI layer uses enterprise documents to process requests, provide answers or recommended actions, and identifies source documents supporting its responses. Jason Mann, VP of IoT at SAS, noted that RAM scales to large data volumes with continuous updates, enabling companies to apply chatbot and conversational AI technologies to corporate knowledge bases, integrate AI-powered knowledge services into applications through APIs, and support AI agent development.

The platform addresses implementation challenges organizations face when deploying generative AI and large language models with enterprise data systems. RAM provides integration capabilities for AI applications ranging from chatbots to automated agents within existing system architectures.