SAS Targets Unstructured Data with RAG Platform
SAS has released Retrieval Agent Manager, a no-code platform designed to extract insights from unstructured enterprise documents using retrieval augmented generation technology.
Built on the retrieval augmented generation (RAG) framework, the platform RAM ingests and processes unstructured documents and evaluates and selects the best configurations for rapid interaction with those documents via an API or chatbot.
RAM also supports the plug-and-play use of GenAI services such as LLMs and vector databases. And it adds an agentic AI layer to automate complex workflows grounded with enterprise data.
SAS positions RAM for regulated industries where document retrieval affects compliance outcomes. The vendor cites anti-money laundering investigations, where fraud teams must rapidly access transaction records and regulatory documentation.
Through its agentic AI layer, RAM uses the company's own data and documents to understand requests, show relevant answers or recommended actions, and transparently share the source documents and data on which its answers and recommendations are based.
The new SAS solution does not use enterprise data to train or fine-tune an LLM. Rather, it keeps the data and LLM separate, creating a knowledge service that brings together the corporate data and LLM at the right time to generate a relevant and timely response.
Insurance claims adjusters could retrieve policy language and prior claim files, according to SAS. Public sector agencies might surface answers from policy manuals and case archives to reduce citizen wait times.
The healthcare use case involves synthesising patient notes and clinical protocols while maintaining HIPAA compliance, though SAS provided no deployment examples.
The company states RAM complements predictive maintenance systems by navigating legacy manuals, inspection reports and field service bulletins. When IoT sensors detect equipment issues, the system retrieves relevant repair documentation.
"SAS Retrieval Agent Manager can scale to very large data volumes that are updated continually," said Jason Mann, VP of IoT at SAS.
"RAM makes it easier for a company to apply technologies like chatbots and conversational AI to its corporate knowledge base, integrate GenAI-powered knowledge services into existing applications via robust APIs, and support the development of AI agents."