AI Agents
Goal-driven assistants that reason over your data, make recommendations, and take actions inside Fusion — not just generate text.
What it is
An AI agent combines GenAI with two new things: a goal and tools. Instead of waiting for you to ask, an agent has a job — "advise on this deal," "screen this candidate," "reconcile this invoice" — and it can read Fusion data, call APIs, and propose or take actions.
The pattern: you give it a goal, it reasons over the data it can see, and it produces a recommendation (and often a draft action you approve with one click).
What makes a Fusion agent different from a copilot
A copilot waits for you to ask. An agent runs proactively in the workflow. The Deal Advisor doesn't wait for a sales rep to type a question — it watches the opportunity and surfaces risk when something changes.
Fusion agents are also system-native: they have the same permissions you do, they respect approval hierarchies, and their actions appear in the audit log just like a human's would.
Examples by pillar
- ERP — Invoice reconciliation, Expense Receipt agent, Cash Forecast agent.
- HCM — Manager Concierge, Talent Advisor, Team Sync Advisor, Candidate Screening agent.
- EPM — Planning Agent (answers ad-hoc finance questions without building a report).
- CX — Deal Advisor, Account Product Fit Agent, Escalation Prediction Agent.
22+ embedded agents are now available across Fusion ERP, HCM, SCM, and CX, with more added each quarterly release.
How you build new ones
Oracle ships embedded agents as part of the regular quarterly Fusion release — you turn them on. For agents Oracle doesn't ship, AI Agent Studio (Release 25C+) lets you and your partners configure custom agents with no-code templates. We cover both paths in Enablement.
Action checklist
Tap each step as you complete it.