Clustering Skill

This skill lets the AI Assistant group similar data points into clusters directly from chat. Use this skill when you want to identify segments in your data, compare how those segments differ, and optionally receive meaningful labels and suggested actions for each segment. The skill uses the same clustering configuration that is available in Analytical Designer.

How It Works

The Clustering skill runs cluster analysis on your data and returns the result in the AI Assistant. When activated, the assistant:

1. identifies the metrics, attributes, and clustering parameters from the user request
2. uses the existing visualization flow to prepare the data
3. runs clustering on the selected data
4. embeds the clustering visualization in the chat
5. if data sharing is enabled, labels the clusters, summarizes the differences between them, and suggests targeted actions for each segment

The skill is designed for users who want to explore natural segments in their data without leaving the assistant. It can support both modes:

• orchestration only, when data sharing is disabled
• orchestration plus interpretation, when data sharing is enabled

Data Sharing

The Clustering skill always respects the AI Assistant data sharing setting.

When data sharing is disabled, the assistant can still run clustering and display the visualization, but it does not use clustering results for natural language interpretation.

When data sharing is enabled, the assistant can also interpret the clustering output, assign meaningful labels to segments, summarize the main differences between them, and suggest targeted follow-up actions.

Examples

Run Clustering from Chat
The user asks: Are there significant groups of products by sales?
The AI Assistant then:

• identifies the relevant data for the request,
• runs clustering on the selected input,
• shows the clustering visualization in the chat, and
• if data sharing is enabled, summarizes the resulting segments.

Find Customer Segments
The user asks: Find natural segments in my customers based on revenue and order frequency.
The AI Assistant then:

• selects the requested metrics,
• runs clustering,
• displays the scatter plot with the resulting groups, and
• if allowed by data sharing settings, labels the segments and explains what makes them different.

Label Clusters and Suggest Actions
The user asks: Group customers into segments and tell me what to do with each one.
The AI Assistant then:

• runs clustering on the relevant customer data,
• shows the result as a clustering visualization,
• labels the clusters in business terms, such as high-value, at-risk, or growth opportunity, and
• if data sharing is enabled, suggests targeted actions for each segment.

More Example Prompts

• Cluster our stores by performance.
• Find similar customer profiles.
• Group products by sales and margin.
• Show natural segments in regional performance.

Limitations

• Clustering is available for scatter plots only
• The skill uses the BIRCH clustering algorithm
• You must define the number of clusters
• The threshold parameter must be greater than 0 and smaller than 1
• A threshold closer to 0 produces more numerous, smaller clusters
• A threshold closer to 1 produces fewer, larger clusters
• Clustering works best when the selected data has enough points and meaningful variation
• Natural language interpretation of clustering results is available only when data sharing is enabled
• Cluster quality depends on the selected metrics, attributes, and the distribution of the underlying data

Error Handling

If clustering cannot be created or does not produce a useful result, the assistant explains the issue and suggests a next step.

Examples include:

• too few data points for clustering
• input data that does not fit the clustering requirements
• clusters with low separation or very similar centroids
• a request that does not match supported clustering inputs

In these cases, the assistant may suggest changing the selected metrics, adjusting the number of clusters, changing the threshold, or filtering the data differently.