Overview of AI Agents and Agentic Apps

Use AI agents and agentic apps in Oracle Permitting and Licensing to help agency staff, inspectors, and public users complete work more efficiently. AI functionality can summarize inspection information, answer questions about application records, generate communication text, review submitted information, monitor workload trends, and recommend next actions.

Oracle delivers AI agents for Public Sector Compliance and Regulation in AI Agent Studio under the PSC family. Agencies can use delivered agents as provided, copy delivered templates to create agency-specific agents, or configure AI functionality that’s embedded directly in application pages and workflows. Some agents use Retrieval-Augmented Generation (RAG) documents that agencies provide to support responses with agency-specific reference content.

You can use the topics in this chapter for Public Sector-specific implementation guidance, including offering-level AI setup, security considerations, page integration, and setup for delivered AI experiences in Oracle Permitting and Licensing.

For general information about using AI Agent Studio, see Oracle AI for Fusion Applications: How do I use AI Agent Studio?

For information about access requirements for AI Agent Studio, see Provide Access to Configure AI Agents in Oracle Permitting and Licensing.

How AI Functionality Is Implemented

AI functionality in Oracle Permitting and Licensing can be delivered in several ways. The setup you complete depends on how users access the AI experience and whether the agent is embedded, template-based, scheduled, invoked by configuration, or part of an agentic app.

AI Functionality Type	Description	Typical Setup
Embedded AI functionality	AI is part of an existing page, drawer, or business flow. Users don’t copy a template or add the agent to a page.	Enable AI functionality for the applicable offering, configure security, and verify the experience.
Template-based AI agent	A delivered template is copied or used in AI Agent Studio, then configured for your agency.	Enable AI functionality, configure security, copy or use the template, publish the agent, and connect it to a page, tile, or other access point.
Page-integrated AI agent	An AI Assist button or similar control is added to an extensible application page.	Configure the agent in AI Agent Studio, then use Visual Builder Studio to associate the page or fragment with the agent code.
Runnable or workflow agent	A delivered workflow agent runs from a configured trigger, button, page action, or Business Rules Framework event.	Enable AI functionality, configure security, and complete the product-specific setup for the trigger or action.
Scheduled AI agent	A workflow agent runs on a schedule and produces output such as email notifications.	Configure the agent in AI Agent Studio, set runtime values, configure recipients, schedule the workflow, and publish the agent.
Agentic app	A task-focused application experience coordinates multiple workflow agents and presents analysis, recommendations, and actions in one workspace.	Enable AI functionality, configure security, and verify access to the workspace.

For example, the Permit Health Monitor uses AI Agent Studio, OTBI permit metrics, runtime values, email nodes, and a schedule to send proactive permit workload notifications. The Code Enforcement Reported Incident Review Assistant is a runnable agent that can be invoked through a Business Rules Framework action.

Agentic Apps

Agentic apps bring multiple AI workflow agents together in a single application experience. Instead of opening separate records or running separate agents, users work from one workspace that presents summaries, recommendations, analytics, and action controls.

The Code Enforcement Incident Workspace is an example of an agentic app. It helps code enforcement users review reported incidents, evaluate AI-generated recommendations, view active incident insights, and take follow-up actions from one workspace. The workspace uses multiple workflow agents, including agents for incoming incident evaluation, active incident summaries, and inspection scheduling. The user remains in control of the action; the app can recommend and prepare an action, but the user initiates it.