Anomaly Detection Skill

This skill analyzes data to find anomalies, outliers, and unusual patterns. Use this skill when users ask questions like “are there any unusual patterns”, “find anomalies in my data”, “what looks abnormal”, or “detect outliers”. This is for immediate, ad-hoc analysis, not for setting up recurring alerts or notifications.

How It Works

The Anomaly Detection skill analyzes your data to identify statistical anomalies, outliers, and unusual patterns. When activated, the Assistant:

1. Uses the Visualization skill to create an appropriate chart or table
2. Executes anomaly detection algorithms on the data
3. Identifies data points that deviate significantly from expected patterns
4. Provides visualizations highlighting the anomalies
5. With data sharing enabled, offers natural language summaries explaining what was found

The skill performs immediate analysis on your current data. It’s designed for ad-hoc investigation, not ongoing monitoring. For recurring anomaly alerts, use the Alert skill instead.

Examples

Finding Anomalies in Sales

The user asks: Are there any anomalies in our sales data this quarter? The AI Assistant then:

• creates a visualization for the relevant sales metric over the quarter,
• runs anomaly detection, and
• highlights unusual points and explains what looks unexpected.

Detecting Outliers

The user asks: Find outliers in customer acquisition costs. The AI Assistant then:

• analyzes the metric distribution,
• identifies outliers, and
• visualizes the results so the outliers are easy to spot.

Identifying Unusual Patterns

The user asks: What looks abnormal in website traffic? The AI Assistant then:

• runs anomaly detection on the traffic metric,
• highlights spikes and drops, and
• summarizes what changed and when.

Limitations

• Requires the Visualization skill as a dependency
• Provides immediate analysis only—not for ongoing monitoring
• Anomaly detection algorithms work best with sufficient historical data
• Results interpretation requires data sharing to be enabled