Wednesday, November 18, 2026

Oracle Fusion AI Agent Studio

 If you work with Oracle Fusion Applications, you have probably noticed that AI agents are showing up everywhere, from HR and payroll to finance and supply chain. The tool behind many of them, and the one you use to build your own, is Oracle AI Agent Studio.

This is the first post in a short series. Here I cover what AI Agent Studio is and what it gives you. In the next post I will walk through a hands-on example: embedding a configured agent in a Redwood page using a Guided Journey.

First, what is an AI agent?

An AI agent goes beyond a chatbot that answers questions. Oracle describes an agent as a system that can do three things: observe its environment, data or users; plan the steps needed to reach a goal using large language models; and act through internal and external systems to get the work done.

That means an agent can handle a multistep task with little human help. Typical examples are processing a job offer or scheduling interviews, where the agent has to look things up, decide what to do next and then take action in the application.

What is AI Agent Studio?

AI Agent Studio is a platform built into Oracle Fusion Applications for creating, configuring, testing and deploying AI agents. It lets you customize the agents Oracle delivers, or build new ones, either as single agents or as teams of agents working together.

Because it lives inside Fusion Applications, there is nothing external to install. Oracle's documentation places it under Home > Tools > AI Agent Studio, though the exact navigation can vary by release and by the roles you have been granted.

The main building blocks

Building block

What it does

Agent templates

Prebuilt agents and agent teams for common front- and back-office use cases. Use them as delivered or customize them. They come with access to Fusion business objects, rules, APIs and data.

Topics

Reusable instructions that tell an agent how to handle a common type of request.

Instructions

The rules, role and logic that shape how the agent behaves.

Tools

Connections that let the agent take action, such as reading Fusion data, calling REST APIs or reaching third-party services.

Multi-agent orchestration

Combine agents into teams that coordinate a longer workflow, with human approval for sensitive steps.

Testing and monitoring

Test an agent before production, review its reasoning path, and monitor accuracy, latency and efficiency.

Why it matters for Fusion HCM and Payroll teams

  • It respects your security model. Agents follow Fusion security policies and role-based access, so users only see data their roles allow. This is a key point to explain to any HR or payroll stakeholder.
  • You can start from something that works. Delivered templates give you a working baseline, so you customize instead of starting from a blank page.
  • Humans stay in control. Approvals can be required before an agent performs a sensitive action, which is important for compensation, payroll and other regulated data.
  • You can choose the model. Oracle provides optimized LLMs, and you can bring other models and select them agent by agent.
  • It reaches users where they work. Agents can be used from Fusion Applications, and they can be surfaced on Redwood pages, which is exactly what the next post covers.

Licensing and access

When Oracle announced the studio in 2025, it stated that AI Agent Studio is included with Fusion Cloud ERP, HCM, SCM and CX subscriptions at no additional cost. Licensing terms and feature availability can change between releases, so confirm the current position with Oracle or your account team before you commit a client to a design.

Where to start

  1. Open AI Agent Studio in your environment and look at the delivered agent templates for your product area.
  2. Pick one small, low-risk use case, such as an advisor-style agent that answers questions from existing data.
  3. Configure it, test it, and review how it reasons before anyone else uses it.
  4. Then decide how users will reach it: AI Chat, a Redwood page, or both.

What's next

In the next post I will show how to take a configured/customized/seeded agent and embed it in a Redwood page using a Guided Journey, step by step with screenshots. 

References

No comments:

Post a Comment