What’s New in Power Pilot Gen2

Power Pilot Gen2 expands the platform’s capabilities by supporting new AI models, including Gemini and Claude, in addition to existing models.

This new version also introduces features that improve control, performance, governance, and overall user experience.

Assistant Control

You can now enable or disable an assistant, providing greater control over its availability within the environment.

Data Security and Governance

Exclude hidden items (enabled by default): Tables, columns, and measures marked as hidden in Power BI are not sent to LLMs, ensuring greater control over the data exposed to the model.

Performance and Cost Optimization

Prompt caching (enabled by default): Reuses the dataset schema across queries using the AI provider’s native cache, reducing token consumption by up to 90%.

User Experience

AI-powered question suggestions: After each response, the assistant suggests up to three related questions based on the conversation context and data model. It uses a lightweight (more cost-efficient) model, with fallback to the main model when necessary.

AI-powered page analysis: When enabled, a button is displayed in the widget that sends data from the current report page visuals to the main model. The response is a text-based analysis including:

  • trends

  • anomalies

  • correlations

  • risks

Power Pilot resizing: You can now resize the Power Pilot window within the portal, allowing you to adjust the viewing space as needed. This improves usability, especially during longer analyses or when simultaneously viewing both the report and the assistant.

AI Behavior Customization

System Prompt (Instructions): Allows you to define tone, persona, and business rules to guide responses in natural language. These instructions do not alter internal DAX generation rules or system formatting rules.

Power Pilot Model

Uses a base model from Power Tuning, optimized for Power Embedded.

Includes a free model with a limit of 100 questions per organization.

When this option is enabled, the selection fields for other models are disabled.

DAX Query Timeout

Allows you to configure the maximum execution time (in seconds) for DAX queries.

A value of 0 means no time limit.

In case of a timeout, the AI attempts to rewrite the query in a more performant way before returning an error.

Atualizado