LlamaIndex has launched Agent Document Workflow (ADW), a new architecture designed to enhance AI agents’ decision-making capabilities beyond traditional retrieval-augmented generation (RAG). According to reporting by Emilia David for VentureBeat, the system enables agents to manage complex workflows and make informed decisions based on document content. ADW treats documents as part of broader business processes, maintaining state across steps and coordinating different components. The system can be applied to tasks like contract reviews, where agents extract information, cross-reference requirements, and generate recommendations. LlamaIndex’s solution combines LlamaCloud parsing capabilities with agents to understand context and drive multi-step processes. The company argues that while RAG is important, it remains primitive for enterprises requiring more sophisticated decision-making capabilities.