Tuesday, October 13, 2026

Oracle Fusion AI Agent Studio: Publishing and Embedding the Agents

In my previous blogs, I covered the details on enabling the Fusion AI agent Studio platform as well as AI Agent migration process. In this blog, let's see the step-by-step process to deploy or publish the AI Agents as well as to embed the agents in Redwood pages in Fusion Cloud.


Publishing the AI Agents


Following would be the primary method of publishing an AI Agent in Fusion AI Agent Studio. The published agents can be accessed from AI Chat option located under Me menu.

- Navigate to AI Agent Studio



- Go to Agent Teams tab and navigate to Draft tab

- Edit the
in AI Agent to be published.


- Navigate to Agent Team Settings

- Go to Security tab and click Add -> Add the desired role that should have access to this agent.

- Ensure the above role has the runtime duty privilege.



- Click Update to save the settings.

- Click Publish button


- Confirm the publish action in the popup



- Once published, we should see the confirmation message




- To access the published agent, navigate to Home

- Navigate to Me -> AI Chat



- Navigate to Explore tab.

- We should see our published agent here.



- Click on the agent to start interacting with it.





Embedding the AI Agents in Redwood pages using Guided Journey


Additionally, once an AI agent is published, we can embed it via a guided journey across any redwood page. Today, the guided journeys are supported in almost 90% of transaction pages.

Let's see the steps to do this.

- Navigate to My Client Groups



- Click Show More and locate Journeys Setup and open it



- Open Guided Journey



- Let's create a new Guided Journey for our agent



- Under Tasks section, add a Task of type Agent

- Pick Workflow Agent (do not pick RAG Agent)

- Select the desired workflow agent - 'O2C Conversational AI ERP Assistant' in our case



- Activate the journey and copy the agent code from the guided journey preview.



- Navigate back to the target page where we want to embed the agent.

- I'm embedding my AI Agent in 'Order Management' home page.

- Click Edit Page in Visual Builder Studio




- This will launch the Visual Builder Studio in a new page






- Under the page properties, search for 'Journey Code' and paste the agent code (from previous step) into the “My Journey Code” property.

- Paste the task code (from previous step) into the "Task Codes" property.



- Click the preview button



- This will open a new page showing the actual redwood page (Order Management homepage in our case), and we should see our AI Agent appearing on the top of the page.



- Navigate back to the Visual Builder tab and click Publish.

- This will us show a popup with the summary of the changes made. Enter a description/message to describe the change being made (this ia mandatory). Click Publish to confirm.



 - In the end, we will be presented with a confirmation message about the successful completion of these configurations.




Let's test it!


- Navigate back to Home

- Go to Order Management



- We should see our custom AI Agent on the top of this page.

- Click the 'Ask Oracle' button.



- This should bring up our custom AI Agent, and we can start interacting with the same.


 

With this, we've covered the typical methods of publishing an AI agent as well as embedding them into the Redwood pages.


Share:

Thursday, October 1, 2026

Oracle Fusion AI Agent Studio: Agent Migration Process

In this blog we will understand the process of migrating AI Agents from one environment to the other in Fusion AI Agent Studio. This process was introduced starting with 26A release and helps us export/import individual AI agents as needed.

I've also covered some of the observations/shortcomings in this process based on my experience. Hope they help the readers.

The steps below describe the Export/Import process for an AI Agent starting from 26A.

Export AI Agent from the Source environment


1. In Source instance:

Navigate to Tools -> AI Agent Studio




Go to AI Agent Studio -> Agent Teams




2. From the Agent Teams tab, locate the agent team you want to export and click the download icon


 
3. While exporting an AI Agent, we need to enter below details:

Partner Name: <Company Name>

Description: Agent description

Partner Image: <Company Logo>

 





In the Preview section we can se how our agent would look like:



4. Once the export details are entered, verify details shown in Preview section and click Export to download the JSON file to your local system. 


5. In above example, the exported file would be 'BENEFITS_POLICY_ADVISOR.json'.



We can open the .JSON file and inspect if it exported all the desired details from the source environment:



Import AI Agent into the Target environment


1. Go to the target environment.

2. Navigate to AI Agent Studio -> Agent Teams

3. Click Import button.

 



4. Select the downloaded JSON file and click Open.


 

5. Once the AI Agent is imported, it can be found under Draft tab. The version of the imported agent gets reset to 1.



6. Use the Edit button to edit the imported agent team.



7. Publish the agent team after making the necessary configurations.



Export/Import Observations


Post-import manual steps (Official list mentioned by Oracle):

- Document tool: Upload and republish the document to ensure its content is correctly processed in the new environment.

- Email tool: Reconfigure the email tool to make sure necessary alerts are generated properly in the new environment.

- External REST tool: Reconfigure authentication settings for external REST API tools.


Additional observations based on my testing:

- This process exports and retains the LLM configuration/assignments for the agent. No reconfiguration needed in this case.




- It doesn't export Security configurations/role assignments for the agent. Needs manual reconfiguration/assignment upon import.



- It exports and retains the Questions (3 custom questions) as expected. No reconfiguration needed in this case.




- The Import process doesn't ask if you want to overwrite, in case the agent already exists. My existing agent version was 7 and upon reimporting it was reset to 1 in the Draft area.

- We need to mention Partner details (Name, agent description and even upload company Logo) during export.

This seems like a good way of branding as the agent shows company name and logo during export.

It’s a good feature that will help end users in identifying our own custom agents.

But I observed that the Company Name and Logo doesn’t appear anywhere upon import, or even after publishing the imported agent.

The branding information is not shown anywhere in AI Agent Studio or on the tiles in AI Chat section.




- Document tool loses the actual document - This is expected since Oracle has already published this behavior.

- All above points apply for Export/Import of a seeded agent as well as a fully custom agent (no difference in behavior).

- All above points apply for Export/Import of a Published agent as well as a Draft agent (no difference in behavior).

- Irrespective of the status of an agent in the source environment (Draft or Published) was, it goes in a Draft status in the target environment upon import.


Share: