How Integrations and AI Agents Fit Together

You use AI agents in Oracle Integration when you need to orchestrate different integrations in an adaptable and flexible automation. You're looking to achieve a complex business goal. AI agents orchestrate integrations as tools to fit your business needs.

Integrations Define a Set Order of Steps

With integrations, you define a specific order of steps. The steps are run the exact same way every single time. You use integrations when you want to control the exact steps and order, and to get enterprise connectivity.

In this diagram, a pre-defined orchestration is represented as a series of sequential steps: Step 1, Step2, and Step 3.

In some cases though, you may have too many possibilities and designing an integration would be too complex. For those cases, you can use AI agents.

Agents Adapt: the Integrations Used Depend on the Goal

An AI agent is a software program that uses a Large Language Model (LLM) to reason without human intervention to achieve a specific goal. AI agents decide which steps to take and in which order, depending on what is happening in the environment. AI agents are more adaptable.

In this diagram, an agent reasons by contacting the LLM and reasons on which step to take: Step 1, Step2, or Step 3.

Integrations are Tools for AI Agents

AI agents use tools to communicate with the world. In Oracle Integration, integrations become agentic AI tools. AI agents use integrations as agentic AI tools to connect to the external world. AI agents determine which integrations to use and in which order to achieve the goal that you define. You create AI agents in projects.

AI Agents Use Thinking Patterns to Reason and Make Decisions

AI agents use specific thinking patterns to reason and make decisions. Patterns that are available by default are:
  • ReAct
  • Plan and Execute

You can also customize the existing thinking patterns or create your own for your AI agents.

Integrations are Discoverable from MCP Clients and Agent Frameworks that Support MCP

You can create AI agents in Oracle Integration, or you can create AI agents in other agent frameworks such as AI Agent Studio for Fusion Applications, Langflow, or others. Discover and use integrations as tools in any third-party agent framework that supports MCP.

Regardless of the agent framework that you use, any integration can be used as an agentic AI tool. You can define which integrations in a projects to expose as tools as an MCP server. Each project becomes an MCP server.

Involve Humans for Approval with Human in the Loop

Human in the loop enables human intervention in key decision areas, when errors occur, and to ensure quality and reliability. Human in the loop ensures people are always in control.

For example, you have an AI agent that automates employee onboarding. If important documents are missing or a background check raises a flag, the AI agent can ask Human Resources for review before continuing.

You use Human in the loop:

  • When approval is needed for an AI agent to execute a particular tool
  • When the AI agent needs approval before doing something else
  • When the AI agent doesn't know what to do. Instead of ending, the AI agent can ask a human what the next steps should be
  • For error handling, when an AI agent calls a tool that results in an error, the agent can tell a human about it who can provide feedback on what to do

Corporate Documents become Knowledge Bases for Agent Actions

Corporate documents are no longer silos and difficult to find and consult. Any corporate document can become a knowledge base that AI agents and humans can query and use as reference. You add documents to a RAG knowledge base that exists within Oracle Integration and query those documents from within an integration, or use the knowledge base as a tool for AI agents.