Intelligent Enterprise Workflow Automation Using Generative Artificial Intelligence and Retrieval-Augmented Generation

Abstract

Enterprise workflow automation has progressed from hard-coded business logic to configurable process engines and robotic process automation (RPA). These approaches work well for structured, predictable tasks but struggle with the unstructured documents, ambiguous requests, and exceptions that make up much of real enterprise work. Generative artificial intelligence (AI), particularly instruction-following large language models (LLMs), can interpret unstructured inputs, reason about what to do next, and produce text, data, and even automation logic. Retrieval-augmented generation (RAG) grounds these capabilities in an organization's own policies, procedures, and records. This article proposes a framework for intelligent enterprise workflow automation that combines configurable workflow orchestration, event-driven triggers, RAG-grounded generative AI, bounded tool use, human oversight, and efficient infrastructure resource allocation. Using a design-oriented approach grounded in ten studies published between 2016 and 2023, it defines six automation patterns, from document intake to workflow authoring and debugging, and illustrates the framework with a procurement invoice exception-handling case. The article argues that generative AI should strengthen deterministic workflows rather than replace them. The workflow engine keeps control of process state, permissions, and audit, while generative AI handles interpretation, judgment support, and exceptions.