Major AI Data Analyst Update: Downloadable Data, Analysis Annotation, and Smarter Insights

In the last few months, we’ve been working to make the AI Data Analyst tool smarter, more transparent, and easier to use. This update introduces key enhancements based on your feedback. Here’s what’s new:

Data Export: Take Your Insights With You

Exporting data from the AI Data Analyst wasn’t available, making it tricky to save or share detailed insights—especially when working with large datasets where only truncated tables were shown.

You can now export data directly from the AI Data Analyst! Whether you’re dealing with large datasets or quick summaries, you can easily download your analysis as a file. This feature keeps your workflow smooth and your insights portable.

Referential Integrity (Annotated Analysis)

Our users wanted better transparency regarding how answers are generated and where the data originates.

We’ve introduced "View Annotated Analysis", a feature that shows you:

• Annotated Python code outlining the steps of the analysis.
• Detailed explanations of data aggregation, preprocessing, and cleaning methods.
• Step-by-step breakdowns of the calculations behind the answers.

This feature is collapsible, so you can toggle it on or off as needed.

Before presenting an answer, the AI writes code and annotates its process, ensuring accuracy and reducing errors through a cutting-edge method called Structured Output Generation. Our research backs this approach as a reliable way to minimize hallucinations.

You can enable this feature in your workspace settings for complete confidence in your data.

Other updates

• File Parsing: Better handling of XLSX and CSV uploads for deeper and more precise insights.
• Date Context Accuracy: Improved handling of dates for calculations like "X months/days before" with transparent answers when certain data is unavailable.
• Streamlined Interactions: Eliminated unnecessary technical messages and resolved abrupt chat ends.
• Enhanced Formatting: Tables, line breaks, and bullet points are now cleaner and more user-friendly.
• Semantic Layer Upgrades: We enhanced our Semantic Layer with more metadata for popular data sources to boost analysis accuracy and prevent hallucinations.

We’re committed to making your experience with the AI Data Analyst even better. Let us know your thoughts and keep sharing your ideas — we’re listening!