By Lalit Reddy, Associate Consultant, Centroid
Oracle AI Agent Studio provides a unified, enterprise-grade framework for designing, configuring, and deploying intelligent AI agents that operate on Oracle Fusion data and business processes. This blog serves as a practical and architectural guide for organizations seeking to design an Oracle Fusion AI Agent that generates a quick executive summary of Vendor Performance based on predefined KPIs.
1. Introduction to Oracle AI Agent Studio
Oracle AI Agent Studio is a low-code platform that enables enterprises to:
- Model business entities as reusable Business Objects
- Build AI Agents that understand context and execute actions
- Orchestrate multiple agents through Agent Teams
- Control agent behavior using structured and governed prompting
2. Business Objects in Oracle AI Agent Studio
2.1 Overview of Business Objects
Business Objects represent enterprise data entities that AI agents can understand, query, and manipulate. They provide a semantic abstraction over underlying Fusion tables, REST APIs, or external services.
2.2 Configuring a Business Object
1. Navigate to Business Objects in the application.
2. Click Create Business Object and provide the following details:
- Family and Product – PRC and Self Service Procurement.
- Name and Description – Enter a meaningful name, e.g., Purchase Order Centroid BO.
- Resource Type – Monolith Resource (Choose the appropriate option based on integration needs).
3. Add Path – Enter the resource path, e.g., /purchaseOrders.
4. Add Functions – Define the operations you want to perform (continue in Step 2).
5. Click Done and Save.
Adding Business Object Functions
Once the Business Object is created, the next step is to define functions (operations) that the agent can perform, such as Get, Get all, creating or updating records.
2.2.1 Adding Operations from Specifications
- Navigate to the Business Object Functions tab of the Business Object.
- Click Add from Specifications. (Prebuild functions)
- From the available operations, select the desired function (e.g., Get, Get All Purchase Orders or Get Purchase Orders Lines, schedules).
2.2.2 Configure Get all Purchase Orders Operation
- Select the “get all Purchase Orders” Operation and click on Done.
- Enter the Name and Description for the operation.
- Added some parameters according to the requirements like q and fields.
- URL – { /fscmRestApi/resources/11.13.18.05/purchaseOrders?q=Status=’Closed’&fields=POHeaderId,OrderNumber,Status,ProcurementBU,OrderNumber,Supplier,SupplierId,Total,Currency,Description,OrderDate,BuyerDisplayName,BillToLocation,ShipToLocationCode&onlyData=true&limit=50}
- Thereis no requirement for a payload or Headers for this Operation.
- Once the configuration is complete, click Done, and then Save
2.2.3 Configure Get Purchase Order Operation
- Within the same Business Object, add another operation by clicking Add from Specifications.
- Select the Get Purchase Order operation.
- Enter name and description for the operation.
- By default, you will have URL parameter {POHeaderId} in URL add a description to the parameters and as per requirement we reduced the fields.
- URL: {/fscmRestApi/resources/11.13.18.05/purchaseOrders/{POHeaderId}?fields=POHeaderId,OrderNumber,Status,ProcurementBU,OrderNumber,Supplier,SupplierId,Total,Currency,Description,OrderDate,BuyerDisplayName,BillToLocation,ShipToLocationCode,lines&onlyData=true }
- Once configuration is complete, click Done, and then Save.
2.2.4: Configure Get Purchase Order using Order Number Operation
- Within the same Business Object, add another operation by clicking Add from Specifications.
- Select the Get Purchase Order operation.
- Enter unique name and description for the operation.
- By default, you will not have URL parameter in URL , so to get the details using Order Number we need to pass it in the Q parameter add in the URL and also add a description in Parameter section.
- URL: { /fscmRestApi/resources/11.13.18.05/purchaseOrders?q=OrderNumber='{OrderNumber}’&q=Status=’Closed’&fields=POHeaderId,OrderNumber,Status,ProcurementBU,OrderNumber,Supplier,SupplierId,Total,Currency,Description,OrderDate,BuyerDisplayName,BillToLocation,ShipToLocationCode,lines&onlyData=true&limit=100 }
- Once configuration is complete, click Done, and then Save.
2.2.5: Configure Get Purchase Order Lines & Schedules Operation
- Within the same Business Object, add another operation by clicking Add from Specifications.
- Select the Get all lines operation in the line operation under PurchaseOrders.
- Enter unique name and description for the operation.
- By default, you will have URL parameter {OrderHeaderId}, so to get the details of schedules for each corresponding line we used “expand” query parameter.
- URL: { /fscmRestApi/resources/11.13.18.05/purchaseOrders/{OrderHeaderId}/child/lines?fields=POLineId,LineNumber,Quantity,schedules&expand=schedules&onlyData=true&limit=50}
- Once configuration is complete, click Done, and then Save.
2.3 Configuring the Deep Links:
- Navigate to AI Agents Studio → Deep Links and Create New.
- Click Add to create a new deep link definition select the desired family & Product.
- Enter a unique Name {Purchase Orders Deep Link Centroid} and Description for the deep link.
- In the Relative URL field, enter:
- URL: { /fscmUI/redwood/purchase-orders/view?intent=POUser&poHeaderId={phid} }
- In the Message field, enter:
Click {deepLink} for details. - In the Parameters section, click Add Row and configure:
-
- Parameter Name: phid
- Description: PO Header id
- Data Type: Integer
- (Optional) Enter a sample value for testing.
- Verify the URL parameter {phid} correctly maps to the Purchase Order Header ID.
- Once configuration iscomplete, click Done (if applicable) and then Save to activate the deep link.
3. Add the Business Object and Deep Link into a Tool
Business Object:
1. Go to Tools and click Create Tool.
2. Provide the following:
- Tool Type – Select Business Object.
- Tool Name – A valid name match with your module. E.g-Purchase Order Centroid Tool.
- Family, Product, and Description – Procurement, Self Service Procurement, and valid description.
3. Click on the [+ add] to add the Created Business Object to the tool and enable all functions.
4. Save the tool configuration.
Deep Link:
- Go to Tools and click Create Tool.
- Provide the following:
- Tool Type – Select DeepLink.
- Tool Name – A valid name match with your module.
- Family, Product, and Description – Procurement, Self Service Procurement, and valid description.
- Click on the [+ add] to add the Created deeplink to the tool. E.g., Purchase Orders Deep Link Centroid.
- Save the tool configuration.
4. Creating an Agent Team
- Go to Agents Teams → Create New Agent Team. → Enter the details noted in step 2.1.
2.1 Details:
- Agent Name – Vendor Performance AI Agent Centroid
- Agent Code – short identifier (auto populate)
- Family – Procurement,
- Product – Self Service Procurement,
- Type - Supervisor,
- Maximum Interactions – (20) set interaction limit,
- Description – explain agent purpose,
- Tags – Purchase Orders, Orders, Suppliers.
2.2 LLM: GPT-5 mini
Choose any predefined LLM or you can use your own LLM.
2.3 Security: Add roles, who can see and access them.
2.4 Questions: Add any predefined Question and follow up questions
- A.Please provide the On-time delivery performance metric for last 3 PO.
- B.Calculate Vendor Performance for Order Number US165220, Display the supplier, Vendor Performance bucket in HTML table.
- Calculate the Vendor Performance report for last 50 PO, Display the supplier & Vendor Performance bucket in HTML table
- Click the Create button to finalize and create the Agent Team.
- Next, navigate the Agents section from the side navigation panel.
Create a new Worker Agent. In the Details tab, input the following information.
4.1 Details:
- Agent Name – Vendor Performance Agent
- Agent Code – VENDOR_PERFORMANCE_AGENT
- Family – PRC
- Product – Self Service Procurement
- Maximum Interactions – 20 (Defines the maximum conversational turns allowed per session)
- Description – The Vendor Performance AI Agent is an intelligent, analytics-driven digital agent designed to evaluate supplier performance across procurement. It leverages enterprise transactional data—such as: Purchase Orders, Lines, and Schedules—to generate real-time compliance-aware supplier performance insights.
- Agent Persona and Role – Oracle Fusion Supplier Performance Calculator AI Agent.
4.2 LLM: GPT-5 mini. Disable the Summarization Mode.
4.3 Input Variables: Not assigned.
4.4. Output - Not assigned.
- Add the tools created in the early steps [Purchase Order Centroid Tool , PO Deeplink tool Centroid] to the newly created Agent Team.
- Click on save.
- Once all agents are added, click Publish to activate the Agent Team, making it available for developers to test. In Agent-Explore, you can check.
Leverage AI to optimize enterprise workflows
Oracle Fusion AI Agent Studio Release 26A demonstrates how governed AI can be embedded directly into enterprise procurement workflows. By combining structured Business Objects, configurable Tools, role-based security, LLM-driven reasoning, and orchestrated Agent Teams, organizations can operationalize supplier performance analytics without disrupting existing Fusion processes.
The Vendor Performance AI Agent illustrates a repeatable architecture pattern:
- Fusion REST resources exposed as Business Objects
- Deterministic KPI calculations (OTD%, Completeness%, Vendor Bucket)
- Context-aware reasoning through LLM integration
- Secure deployment via Agent Teams
This approach ensures traceability, compliance alignment, and scalability while transforming transactional procurement data into actionable executive insights. As enterprises mature their AI adoption strategy, such governed agent frameworks will become foundational to intelligent decision automation within Oracle Fusion environments.
Accelerate your AI journey with Centroid
Centroid helps enterprises design and deploy production-grade AI Agents in Oracle Fusion AI Agent Studio. From Business Object modeling and REST integration to Agent Team orchestration and KPI-driven intelligence frameworks, our experts deliver scalable, secure, and performance-oriented AI solutions tailored to procurement and enterprise operations. Contact us to learn more and to obtain a complimentary consultation.