Oracle Analytics AI Agents - Building AI Agents in Oracle Analytics

Author: Philip Godfrey

Oracle Analytics Cloud is a powerful tool allowing users, both individuals and businesses alike, to explore and understand their data by utilizing advanced and augmented analytic capabilities.

These augmentations, such as Auto-Insights, Explain and AI Assistant utilise Machine Learning and Artificial Intelligence to provide even more options for business to explore and understand their data.

One of my favourite new features in this space is Oracle Analytics AI Agents, which is the focus of this blog, made available in the Oracle Analytics Cloud January 2026 update.

The previous blog in this series focused on designing Oracle AI Agents, and this blog will focus on building AI Agents in Oracle Analytics Cloud.

 

What is Oracle Analytics AI Agents?

AI Agents in Oracle Analytics is a natural extension to the Oracle Analytics AI Assistant.

Why is this so important? The AI Assistant is excellent, but what it can miss is that extra layer of knowledge and grounding. After all, it can only provide answers based on the data it’s been trained on. Where AI Agents is particularly useful is that it allows the AI Assistant to interpret natural language questions with greater accuracy and deliver more meaningful, context-aware insights.

 

How does Oracle Analytics AI Agents work?

Oracle Analytics AI Agents uses Retrieval-Augmented Generation (RAG) to enhance generative AI with enterprise data, enabling large language models (LLMs) to “search” relevant information before responding.

 

To help build the AI Agent, we need to understand the key elements incorporated:

 

Building an AI Agent

Authors will be required to configure 4 key components to enable the AI Agent to function.

Dataset: The business data required to be utilized by the AI Agent

Supplemental Instructions: Think of this as additional instruction, a place where you can define the logic, expectations, and interpretation rules that guide the Oracle Analytics AI Assistant’s reasoning[1]

First Message: A welcome message displayed each time you interact with the AI Agent.

Knowledge Documents: These are supporting materials to aid your Agent, in making use of Retrieval Augmented Generation (RAG) to ensure answers are context aware.

 

Authors configure the agent once, defining:

  • What data it can access
  • How it should interpret questions
  • What guidance it should follow

End users then interact with a conversational interface that feels intuitive but is carefully governed behind the scenes.

 

Providing Access to an AI Agent

AI Agents must be enabled before they can be accessed.

When you enable an AI agent in a workbook, users with access to the workbook can access and interact with the AI agent. For example, users might use an AI agent within the context of a workbook to obtain quick clarifications, follow-up answers, and drill-down insights within the context of the analytics content they're exploring.


How will users access?

Oracle Analytics AI Agents are available to end users only when dashboards are accessed in Present Mode. When you enable an AI agent in a workbook, users with access to the workbook can access and interact with the AI agent.

 

What happens when you submit a question to the AI Agent[2]?

  • The agent automatically applies additional custom instructions created by the author, adding important context such as domain-specific definitions or guidance on preferred response formats.
  • A RAG process scans the selected enterprise documents configured by the author, enriching your prompt with focused, relevant information. The enhanced query is then processed by the Oracle Analytics AI Assistant, which orchestrates the underlying LLM prompts to deliver a response that best reflects your intent and the available enterprise data.

 

In the below example shared by Oracle Analytics[3], a HR Agent uses its supplemental instruction to provide full names of employees and made use of the knowledge document on HR policy to assist in its response. Meaning responses are tailored to the business domain knowledge it has access to.




[1] https://docs.oracle.com/en/cloud/paas/analytics-cloud/acubi/supplemental-instructions-oracle-analytics-ai-agent.html#GUID-700AE5A0-4710-4ADC-8849-652305FCA9E3

[2] https://docs.oracle.com/en/cloud/paas/analytics-cloud/acubi/oracle-analytics-ai-agents.html

[3] https://youtu.be/4iv9YsXw5Po?list=PL6gBNP-Fr8KVwTfbpjp2E5_jPWfJz0IyY

Comments