LlamaIndex goes beyond RAG to allow agents to make complex decisions

Popular AI orchestration framework LlamaIndex has introduced Agent Document Workflow (ADW), a new architecture that the company says goes beyond retrieval-augmented generation (RAG) processes and increases agent productivity.

As orchestration frameworks continue to improve, this method could provide organizations with a way to improve the decision-making ability of their agents.

According to LlamaIndex, ADW can help agents “manage complex workflows that go beyond simple extract or match.”

Some agent frameworks are based on RAG systems that provide agents with the information they need to complete tasks. However, this method does not allow agents to make decisions based on this information.

LlamaIndex gave some real-world examples of how ADW would work well. For example, during contract reviews, human analysts must extract key information, cross-reference regulatory requirements, identify potential risks, and develop recommendations. When used in this workflow, AI agents ideally follow the same pattern, making decisions based on the documents they read to review the contract and knowledge from other documents.

“ADW addresses these challenges by treating documents as part of broader business processes,” LlamaIndex said in a Blog post. “An ADW system can maintain status across steps, apply business rules, coordinate various components, and take actions based on document content—not just analyzing it.”

LlamaIndex has previously said that while RAG is an important technique, remains primitiveespecially for companies seeking more robust decision-making capabilities using AI.

Understand the context for decision making

LlamaIndex has developed reference architectures that combine its LlamaCloud parsing capabilities with agents. It “builds systems that can understand context, maintain state, and control multi-step processes.”

For this purpose, each workflow has a document that acts as an orchestrator. It can instruct agents to tap LlamaParse to extract information from data, maintain the state of the document context and process, and then retrieve reference material from another knowledge base. From here, agents can begin generating recommendations for the contract review use case or other actionable decisions for various use cases.

“By maintaining status throughout the process, agents can handle complex multi-step workflows that go beyond simple extract or match,” the company said. “This approach allows them to build comprehensive context around the documents they process while coordinating the various system components.”

Different agent frameworks

Agentic orchestration is an emerging area, and many companies are still exploring how agents – or multiple agents – work for them. Orchestration of AI agents and applications may be possible a bigger conversation This year, agents are transitioning from single systems to multi-agent ecosystems.

AI agents are an extension of what RAG offers, namely the ability to find information based on corporate knowledge.

However, as more and more companies start using AI agents, they also want them to take over many of the tasks that human employees do. And for these more complicated use cases, “vanilla” RAG is not enough. One of the advanced approaches that companies have considered is Agent RAGthereby expanding the agents’ knowledge base. Models can decide whether they need to find more information, what tool they will use to retrieve that information, and whether the context they just retrieved is relevant before producing a result.