Oracle is moving away from click-heavy screens to AI-driven conversations. Oracle AI Agents work directly inside applications, understand the business context, follow security and approval rules, and help users get answers and complete tasks without jumping across multiple pages.

Here’s a FAQ covering what Oracle AI Agents can do today and what’s coming next.

1. How do I migrate AI Agents between environments?
25D: Use FSM to export and import published agents
26A: Import individual agent teams directly (simpler migration)

2. How can response time be improved?
Keep prompts simple. Avoid too many documents. Avoid overly complex agent logic. Oracle is improving performance every release.

3. Is ODA still recommended?
No. Seeded ODA intents are being phased out. Oracle recommends AI Agents for new use cases.

4. Can agents call external systems or OIC?
Yes. Agents can call OIC and external REST APIs. SOAP APIs must be wrapped as REST.

5. Can agents be used in Microsoft Teams or Slack?
Microsoft Teams integration is coming in 26A. Slack integration is planned for a future release.

6. Do I need Redwood UI to use AI Agents?
No. Redwood is required only to embed agents inside transaction pages using Guided Journeys. Agents can still be: Accessed from the Explore UI. Used via REST APIs. Used without Redwood UI.

7. Where can AI Agents be accessed or deployed?
AI Agents can be deployed and accessed: From the Explore tab (Me → AI Agent Studio), Through Guided Journeys, Within transaction pages using Smart Actions, Externally via REST APIs (available from 25D)

8. Can agent’s answers be shown in table format?
Yes. You can guide the agent to return table-formatted answers using simple formatting instructions today. Native tables and richer layouts are coming soon.

9. What are Agentic UX widgets?
Agentic UX widgets are upcoming UI elements that allow agents to show:
Tables, Cards, Forms and Lists. This makes information easier to read and allows simple actions directly in chat.

10. Can users upload files into chat?
File upload and multimodal support are expected in 26A.

11. Is chat history saved?
Not yet. Chat history is session-based today. Persistent history is on the roadmap.

12. Why agents responses might differ for the same question?
LLMs are stochastic, so slight variation is expected. Oracle is improving guardrails and prompt guidelines to reduce overreaching behaviour and improve consistency.

13. Are OTBI or BI Publisher supported as agent data sources?
No. OTBI and BI Publisher are not supported as seeded data sources for now.

14. How can I improve agent response times?
Keep prompts concise, Optimize API response sizes, avoid overly complex agent designs, limit the number of large documents.

15. Is there extra cost for custom agents?
Yes. Custom agents incur additional cost when used in production. Agents built from scratch are always custom.