Oracle Analytics
Oracle Analytics Cloud · Generative AI

Unlocking the Gold Mine of
Unstructured Data with OAC AI Functions

Turn free-text columns into strategic, dashboard-ready insights — without writing a single line of code.

Write a prompt Point at text column Dashboard-ready insights
GEN
AI_GENERATE()
Row-level text enrichment & classification
FIL
AI_FILTER()
Meaning-based row classification & labeling
AGG
AI_AGG()
Group-level summarization & synthesis
Start Here Step 1: Setup Guide Step 2: Hands-On Exercise
01

Your Most Valuable Data Is Hiding in
Free-Text Columns

A lot of the most valuable business signals never make it into dashboards — because they live inside free-text columns that traditional BI tools simply can’t read. These columns exist across every industry and every department:

💬 Customer feedback
🎫 Support tickets
📝 Survey comments
⚠️ Incident summaries
Product reviews
📄 Case notes
🛒 Post-purchase surveys
👥 HR feedback
📑 Claims narratives
⚙️ Operational issues
🏥 Discharge summaries
🔬 Research summaries
What this looks like in practice
Before — Raw Text
“Shipping took 3 weeks, ridiculous”
“Love the quality, will buy again”
“Price way too high for what you get”
After — Structured Insights
Negative · At Risk · Shipping
Positive · Retained · Quality
Negative · At Risk · Pricing
02

Three Functions That Read Your Data

OAC’s AI Functions are built directly into the workbook expression editor — right alongside SUM() and RANK(). Write a prompt in plain English, point it at any text column, and the LLM classifies, filters, or summarizes your data in real time.

AI_GENERATE()
Row-Level
Text Enrichment

Classify sentiment, extract complaint categories, label each row with AI-generated insights. One LLM call per row, results appear as a new column.

COMMENTS → “Negative”
AI_FILTER()
Meaning-Based
Row Classification

Ask a plain English question about each row — “Is this customer at risk of leaving?” — and use the result to assign meaningful labels like “At Risk” vs “Retained”, “Fraud” vs “Legitimate”, or any category you define.

COMMENTS → “At Risk” / “Retained”
AI_AGG()
Group-Level
Summarization

Aggregate free-text across a dimension — region, occupation, product line — into a single executive-ready narrative. One summary per group.

250 comments → 1 paragraph
Zero Code. Zero Data Movement. Zero ML Expertise.
All three functions run natively inside OAC’s expression editor. Connect to OpenAI, Anthropic, Google, Cohere, OCI Gen AI, or any supported provider via a Select AI profile. Every expression in this guide has been live-tested on real OAC instances.
03

Who Is This For?

This guide is for OAC practitioners, solutions architects, and data analysts who want to add AI-powered text analysis to their analytics workbooks. If your organization already uses Oracle Analytics Cloud and has free-text columns sitting unused in the data warehouse, these functions let you unlock that data immediately — no additional tools, no data science team required.

Learning Path

Two guided steps — from model registration to a fully AI-powered churn detection dashboard.

1
Setup Guide

Register AI Models & Configure Select AI

Complete walkthrough — OCI Resource Connection, Autonomous AI Lakehouse, credential creation, network ACL, AI profile, and troubleshooting with exact error fixes.

✓ By the end: your OAC instance can talk to any LLM provider
OAC Setup Select AI DBMS_CLOUD_AI
Start Setup
2
Hands-On Exercise

Build It Yourself: Step-by-Step Walkthrough

Click-by-click instructions — sentiment classification, churn risk flagging, group summarization, dashboard assembly. Explore how each AI function transforms your data, then compare against the reference DVA file.

✓ By the end: a complete AI-powered churn detection workbook
Hands-On Lab Workbook DVA File
Start Building

No Python Required

All AI functions work natively in OAC's expression editor — no external code, no data movement.

🔑

Multi-Provider Support

Connect to OpenAI, Anthropic, Google, Cohere, OCI Gen AI, and more via Select AI profiles.

📊

Real Data, Live-Tested

Every expression and output has been validated on real OAC instances with the SH schema.

📦

Reference DVA File

A downloadable OAC workbook (.dva) with the completed hands-on exercise — all AI-powered calculated columns and visualizations pre-built for comparison.

Available alongside these guides for import into your OAC instance.
R

Ravi Shanker B

Principal Solutions Architect · March 2026