Oracle Fusion AI Agents Studio

• Introduction to AI Agents: Understanding the paradigm shift in Oracle Cloud automation.
• Getting Started with Oracle AI Agent Studio: Navigating the 26C Studio, including Applications, Workflows, Resources, Recents, Marketplace, Monitoring, and Evaluation.
• The Anatomy of an Agent: Exploring the core building blocks—LLMs, Workflow Agents, Multiagent nodes, Tools, Topics, Business Objects, variables, and prompts.
• Real-World Application: Deep dive into high-value AI Agent use cases in ERP, HCM, and CX.
• Building in the 26C Environment: Understanding how applications, workflows, resources, agents, tools, and topics are organized before starting the hands-on labs.
• Introduction to Agentic Apps: A conceptual overview of how Agentic Apps combine Information Displays, Ask Oracle advisors, Actions, Communications, and specialized agents into a unified business experience
• Hands-On Lab 1: Rapid Deployment: Creating your first AI Agent using Oracle-delivered Best Practice Templates.

• Understanding AI Agent Security: Concepts of Role-Based Access Control (RBAC) within the Agent framework.
• Security Configuration Blocks: Analyzing policies, user roles, and data access requirements.
• Design-Time vs. Runtime Security: Differentiating Studio developer access, architect access, end-user invocation access, and the data privileges applied when an agent runs.
• Fusion Data Security: Understanding how Oracle Fusion job roles, data roles, and application security govern the records returned through Business Objects and Fusion APIs.
• Hands-On Lab 2: Runtime Security: Setting up security profiles and access controls for End Users from scratch.
• Hands-On Lab 3: Design-Time Security: Configuring specific Pillar-level security for Developers and Architects designing agents.

• Understanding Deep Links: Mechanisms to drive user navigation within Oracle Fusion apps.
• Configuring Deep Links: Step-by-step setup of Deep Links within AI Agent Studio.
• Working with Business Objects: Overview of the Business Object (BO) structure and data availability.
• Business Object Setup: Configuring BOs from scratch for read/write operations.
• User Session Tool: Retrieving signed-in user context such as username, person number, and other identifiers to create personalized and user-aware agent experiences.
• Chat Attachments Reader: Using the Multi-File Processor context to read files uploaded during a conversation, return extracted Markdown or Base64 content, and understand the purpose of the conversation identifier.
• Hands-On Lab 4: End-to-End Build: Creating a Custom AI Agent utilizing the User Session Tool, Business Objects, and Deep Links for personalized data access and seamless user navigation.

• External REST Tool Overview: Extending agent capabilities beyond Oracle Fusion.
• Strategic Integration: Identifying scenarios for External REST API usage versus internal tools.
• Security & Authorization: Configuring Authentication (OAUTH, Basic, API Key) for external connections.
• API Setup: Registering and configuring REST endpoints within the Studio.
• Workflow Triggers: Understanding chat, webhook, email, scheduled, and signal-based workflow invocation and selecting the appropriate trigger for each business process.
• Model Context Protocol (MCP): Introduction to connecting an AI Agent Studio tool to an external MCP server and understanding when MCP is more suitable than a conventional External REST integration.
• Enterprise Connectors: Overview of SharePoint, web-crawler, document, OCI, and Oracle Integration Cloud connectivity options for extending agent knowledge and actions.
• Hands-On Lab 5: Advanced Integration: Building a Custom AI Agent that authenticates, handles errors, and interacts with third-party systems through External REST APIs.

• Architecture Patterns: Overview of Conversational, Workflow, RAG-based, Transactional, and Multiagent patterns in AI Agent Studio.
• Workflow Strategy: Decision matrices for selecting RAG, Transactional, Multiagent, deterministic workflow, or hybrid AI orchestration patterns.
• The Art of Prompt Engineering: Understanding how system prompts establish behavior and guardrails while user prompts provide runtime requests, context, and business inputs.
• Best Practices: Designing tool-aware prompts, reducing hallucinations, decomposing complex requests, grounding responses in retrieved data, and preventing premature answers.
• Multiagent Architecture: Designing supervisor-style orchestration with the Multiagent node, including worker specialization, routing instructions, explicit input mapping, response consolidation, and testing.
• Workflow Nodes and Flow Control: Configuring LLM, Agent, Tool, Set Variables, Code, IF, Loop, While, Wait, Switch, Run in Parallel, Error Handler, Human Approval, Return, and Reference nodes.
• Structured Output Specifications: Designing valid JSON schemas for objects, arrays, nested records, transaction confirmations, and standardized error responses.
• Document Intelligence: Understanding static document RAG, chat attachments, the Document Processor node, and reusable Document Schemas for extracting structured information from invoices, contracts, and business documents.
• Hands-On Lab 6: Optimization Workshop: Applying advanced workflow nodes, reusable logic, structured outputs, and prompt engineering techniques to improve workflow reliability and agent responses.

• Troubleshooting: How to trace AI Agent and Workflow execution, inspect node inputs and outputs, analyze variables, review logs, and isolate tool or prompt failures.
• Targeted Node Testing: Testing individual LLM and Agent nodes, refining instructions and model parameters, and comparing results before running the complete workflow.
• Workflow Debugger: Tracing execution paths with breakpoints, stepping through nodes, inspecting context, overriding configuration, pinning node output, and rerunning from a selected node
• Introduction to Guided Journeys: Embedding agents into user workflows.
• Deployment Strategies: Configuring agents within Guided Journeys for specific UI pages.
• Environment Promotion: Best practices for promoting agents and workflow artifacts between Development, Test, Stage, and Production environments and revalidating dependencies after deployment.
• Packaging and Dependency Considerations: Understanding the deployment and packaging options available in the 26C environment, artifact dependencies, and validation checks required before promotion.
• Hands-On Lab 7: Go-Live Simulation: Testing, debugging, publishing, and deploying a fully functional AI Agent through Guided Journeys.

• Monitoring Options: Utilizing dashboards and activity views to track agent usage, workflow status, latency, tool and node execution, trigger activity, and error rates.
• Evaluation Frameworks: Methodologies and repeatable test cases to assess Agent correctness, relevance, completeness, groundedness, tool-selection accuracy, and safety.
• Continuous Improvement: Feedback loops for refining agent performance over time.
• Production Governance: Applying auditability, human approvals, access reviews, and responsible AI oversight to production deployments.
• Hands-On Lab 8: Operations Command: Reviewing workflow activities, identifying a failed node, applying a correction, rerunning the workflow, and performing correctness and relevance evaluations.

1. Pay Analyst AI Agent (Template-Based Deployment)

This hands-on demo shows how to rapidly deploy an AI agent by starting from an Oracle-delivered template and tailoring it to your organization’s needs. Participants will create a copy of the seeded Pay Analyst agent and preview it end-to-end, demonstrating how employees can ask questions and quickly understand their payslip details, deductions, and benefits in plain language.

Key capabilities you’ll configure and demonstrate:

• Template-based AI agent deployment: Clone a delivered agent, configure it, and make it ready for business users.

• Employee-friendly payroll explanations: Guide the agent to provide easy-to-understand summaries of earnings, deductions, and benefits.

• Tour of AI Agent components: Walk through the building blocks (what matters, what to change, and how to validate outcomes).

2. Procurement Automation AI Agent (Built from Scratch)

In this project, you will learn how to build a real business automation AI agent from the ground up. The agent collects a few guided inputs from the requester and automates the procurement flow—reducing manual entry, improving data quality, and accelerating the
path from requisition creation to downstream purchasing orders.

Key capabilities you’ll build and demonstrate:

• Guided intake for procurement requests: Capture what the user needs in a structured conversation.
• Business Objects + Fusion API orchestration: Create multiple business objects to read, create, and update Oracle Fusion data seamlessly.
• User-aware automation: Leverage delivered tools to identify the user and apply preferences to speed up requisition creation (less typing, fewer errors).
• End-to-end process experience: Demonstrate how AI agents can drive a complete business process using Fusion APIs—without needing users to navigate multiple screens.
• Custom Agent Development: Gain the expertise to build AI agents from the ground up, utilizing Fusion APIs to orchestrate complex procurement tasks.

3. Manager 360 Ops Insight AI Agent (Multi-Agent Supervisor Pattern)

This project teaches how to design a Supervisor AI Agent that provides managers a 360° overview of team responsibilities, workload, and operational activities—by orchestrating multiple specialised agents behind the scenes. It’s a practical, executive-friendly use case that shows how multi-agent design scales across teams and processes.

Key capabilities you’ll build and demonstrate:

• Supervisor agent orchestration: Design a “manager brain” agent that routes tasks to multiple supporting agents each focused on a specific operational area.
• Multi-Agent Team Deployment: Learn the architecture of building and deploying multi-agent teams, enabling parallel processing of complex management activities
• 360-Degree Team Insights: Provide managers with comprehensive visibility into team performance and activities, consolidating data into a single operational view
• 3rd-party intelligence integration: Build agents that connect to external tools (example: incident/ITSM systems) and apply intelligence to the fetched data for trends, risks, and priority calls.

4. Supplier Sync Workflow AI Agent (Workflow Automation + Controls)

In this project, participants create Workflow AI Agents from scratch to keep supplier data synchronised between Oracle ERP and a third-party application. The focus is on real-world workflow engineering: conditions, variables, validations, updates, exception handling, and notifications—so the solution is production-minded, not just a demo.

Key capabilities you’ll build and demonstrate:

• Supplier sync orchestration: Ensure supplier master data stays consistent across systems, reducing downstream payment, compliance, and onboarding issues.
• Workflow node configuration: Implement workflow nodes, apply conditional logic, set and pass variables, and drive outcomes based on business rules.
• Fusion Data Management: Create business objects specifically designed to update and maintain data records within Oracle Fusion applications based on workflow triggers.
• Automated Notifications: Enhance process visibility by configuring the workflow AI agent to send automated notifications to relevant stakeholders upon successful synchronisation or status changes

5. Intelligent Invoice Validation/Creation Using Workflow AI Agent in Oracle Fusion

In this project, you will learn how to build a real business automation workflow AI agent from the ground up. The agent receives invoice data from an external application through a webhook, processes an array of invoice lines, validates referenced purchase orders against Oracle Fusion, and intelligently decides whether to create the invoice or return a clear exception summary. This use case demonstrates how workflow AI agents can orchestrate structured enterprise processes with looping, branching, and system interaction.

Key capabilities you’ll build and demonstrate:

• Webhook-based invoice intake: Receive invoice header and line details from an external application in a structured payload.
• Array handling with Loop node: Process multiple invoice lines dynamically by iterating through each line in the payload.
• PO validation against Oracle Fusion: Check whether each referenced purchase order exists in Fusion before allowing invoice creation.
• Switch-based exception routing: Route each invoice line into business outcomes such as valid PO, missing PO, duplicate PO, or invalid/blank PO.
• Exception-first processing: Prevent invoice creation when one or more lines fail validation and generate a meaningful exception summary for the calling system.
• Conditional invoice creation: Automatically create the invoice in Oracle Fusion only when all required validations pass successfully.
• Workflow orchestration from scratch: Gain hands-on expertise in building Workflow AI Agents using trigger inputs, variables, Loop, Switch, and downstream system actions.
• Real-world integration design: Demonstrate how Oracle Fusion AI Agent Studio can orchestrate deterministic business processes across external applications and Fusion APIs without requiring users to manually navigate multiple systems.