Abstract

The current landscape of Enterprise AI is rapidly evolving, shifting from just having a dialogue with a large language model (LLM) to actually having sophisticated, goal-oriented Agentic Workflows. This shift is driven by the need to not just generate responses from an LLM but to also to have an AI do tasks based on the information retrieved. Most Enterprises have a vast amount of data in PDFs or other documents that can be used to ask complex questions, Retrieval Augmented Generation (RAG) can help answer questions within Enterprise data, however if there are more complex questions that require private as well as Public Internet data, we need an agent that can fetch data from both sources, reflect on it and provide detailed reports. In this paper, we look at the NVIDIA AI-Q Research Agent blueprint, an agentic system that can generate detailed reports based on both internal and external data. We walk through how to deploy, how to scale and provide sizing guidance.