AI Agent Studio provides new tools for automation

Key takeaways 

• Oracle AI Agent Studio is the AI development platform for Fusion Applications that lets you build, test, and deploy AI agents to automate more of your business.
• It includes hundreds of templates from Oracle and its partners to give you a head start on deploying AI-driven automation. 
• New observability and testing features help build trust and oversight of agent development and deployment. 

Oracle AI Agent Studio is the AI development platform for Fusion Applications that lets you build, test, and deploy AI agents to automate more of your business. We introduced AI Agent Studio earlier, and now we’re sharing an updated demo that illustrates some of the most important product advancements. These include an expansion of its standards-driven ecosystem, the introduction of workflow-enabled agents, and new tools for measurement and evaluation.  

What you’ll see 

The demo follows Marie, a power user in a shared service center who wants to help employees automate self-service requisitioning. She begins by reviewing agent templates in AI Agent Studio. These include templates delivered by partners as part of the new AI Agent Marketplace, the trusted source for partner-developed, Oracle-certified AI agents. Next, Marie notices important controls that she can use to guide agents, limit their scope where appropriate, and provide access to third-party tools.  

Marie selects the Quote-to-Purchase Requisition Assistant from the list of available templates. It is a workflow-enabled agent team designed to convert supplier quote documents into purchase requisitions in Oracle Self Service Procurement. She sees a clear, graphical representation of the workflow and reviews how it ingests supplier quotes, analyzes them, and calls on specific Fusion Apps data and functions to create a purchase requisition.  

The template is a good starting point, but Marie decides to customize the workflow by adding an explicit approval step (i.e., human in the loop) and removing the now superfluous email notification. After making these changes, she publishes the agent. 

You’ll then see the agent in action as an end user named Eric uses it to upload a PDF document detailing a supplier quote. The agent processes the document and then asks for confirmation from Eric before creating the requisition.   

Why it matters 

This demo shows new or significantly expanded capabilities in AI Agent Studio, including:  

• Open, standards-based ecosystem: AI Agent Studio is part of Fusion Applications, and it works with non-Fusion applications, too. It supports open standards such as Model Context Protocol (MCP) and Agent2Agent Protocol (A2A) , and REST APIs. And while it comes with access to LLMs by OpenAI, Meta, and Cohere, you can choose others to best suit your needs (i.e., LLMs by Anthropic, Google, xAI, etc.).  
• AI Agent Marketplace: Offered from within AI Agent Studio, this is the source for partner-delivered agent templates. Oracle tests, approves, and supports partner templates. 
• Workflow-enabled agents: AI Agent Studio now supports agents that follow a specific, defined workflow. The addition of these deterministic agents may be the most important functional advancement on this list. Each one combines branching logic, triggers, and nodes that connect worker agents, apps, LLMs, and vector databases, allowing multiple agents to work together to accomplish a goal following a clear, step-by-step execution plan. This means that they can generate the same output from the same input (unlike probabilistic agents, whose output varies even with the same input), and also use reasoning capabilities to complete tasks (e.g., this number isn’t labeled but looks like a SKU, so the agent treats it as such).  
• Explainable and measurable AI: In the context of enterprise AI, governance and trust are key. AI Agent Studio features a testing framework that provides an expanded set of tools for evaluating semantic accuracy (i.e., outputs that correctly represent the underlying facts and context), tracing (i.e., what steps an agent takes), and agent performance (i.e., speed, accuracy, etc.). It includes features like dataset management, A/B comparisons, performance monitoring, token usage, debugging, and agent-level guardrails.  

Summing up 

AI Agent Studio is an important part of Oracle’s AI strategy and for automation in Fusion Applications, and it is included with your Fusion Applications subscription. We’re growing its standards-driven ecosystem, expanding capabilities with workflow-enabled agents, and providing you with new tools for measurement and evaluation. Our goal is to help developers, application administrators, and power (or super) users bring agent-driven automation to your business. For more details, see “Oracle AI Agents for Fusion Applications.”