Release Notes - 2026.02.04

Welcome to the latest One Model product release update. This article provides an overview of the  product innovations and improvements to be delivered on 4 February 2026.

One AI

One AI Assistant has smarter, agentic metric and dimension matching: One AI Assistant is now significantly better at understanding how people actually ask questions. You no longer need to use the exact metric or dimension name configured in your model. This improvement makes One AI feel far more natural and intuitive for stakeholders, while still respecting permissions and configuration. You can now ask questions like:

• “How do promotions compare to transfers in 2023?”
• “What’s our quit rate vs. fire rate?” 
• “How many people work here?” instead of “headcount”
• “How many people work for [manager name]?”
  • Configuration still matters—if multiple dimensions contain identical nodes, the Assistant may not know which one to use. 
  • Agentic search for dimension nodes is coming soon.

Prompt routing now shows multiple options: One AI Assistant now offers multiple routing options in AI Mode when a prompt could apply to more than one capability. For example, typing “headcount” might mean you want a visualization (like a key value tile), Storyboards related to headcount, or an Answer about headcount. Instead of guessing, the Assistant will present the relevant options so you can choose the right one. This also prevents Answers from overpowering prompts that contain matching keywords. (ref ON-4128)

• Automatic routing to Chat or Analyze from AI Mode: If you’re in AI Mode and your prompt is clearly intended for Chat or Analyze, One AI Assistant can route you automatically without manual mode switching. This makes the experience more seamless by reducing extra clicks and manual mode switching. (ref ON-4128
• Faster, more concise written responses for One AI Assistant Answers: We improved the speed and conciseness of the natural language written response for One AI Assistant Answers. We updated the model and adjusted the prompt instructions to produce shorter, clearer responses. (ref ON-4064)

User Experience Improvements

• Chart image exports now use the Storyboard Tile title as the downloaded filename. When you export a chart from a tile (for example, “Headcount by Department”), the PNG file will inherit that title (e.g. Headcount by Department.png) instead of a generic chart.png name. If a tile has no title, we fall back to a sensible default name, and any characters that aren’t valid in filenames (such as /, \, : or *) are automatically sanitized so downloads succeed reliably. (ref 26734)
• We have added the following Storyboard events to Audit Logs - Storyboard copied, Storyboard Settings edited, Storyboard Sharing edited, Storyboard deleted, Storyboard modified, Replace Storyboard and Storyboard Viewed. Learn more in our Audit Logs Guide. (ref 24645, 26253)
• We’ve started updating the storyboard permissions terminology in the One Model application so that the “Publish To” option is now labeled “Share With” and "CanPublishDashboards" is now "CanShareStoryboards" wherever it appears. This change aligns the UI language with how users share storyboards with Data Access Roles, while preserving existing behaviour such as selecting/deselecting all roles and controlling which roles can receive shared content. Terminology in the Help Center guides has been updated to reflect this change also. (ref 26197, 26198)
• Fixed an issue where a "SELECT * INTO" statement in SQL Explorer would apply the "Limit" configured on the page, unexpectedly shortening the size of the table that was created. This limit will no longer apply to "SELECT * INTO" statements (ref 26334)

Data Processing Innovations

• We added a Processing Script validation to prevent users from using capital letters in the names of objects they were defining, because objects with capital letters in the name could cause errors in other parts of the application (ref 26585)

Minor Improvements & Bugs Squashed

• Special characters in dimension labels now display correctly in chart tooltips. Previously, labels containing characters such as “<” (for example, “In < Office”) were being truncated in hover tooltips, showing only the text before the special character. We’ve updated the tooltip rendering so that the full label text is shown when hovering over chart elements, consistent with how labels appear elsewhere in the UI. (ref 26663)
• We resolved an issue where selecting both Subtotal and Change Indicator at the same time caused the page to display an error message and fail to load. Charts and tables will now display correctly when using both of these features together (ref 26755)
• For our early beta testers of the KPI Tile with AI Insights: We've deployed a small update to improve visual consistency, which includes ensuring correct branding colors are applied to the Insight text and standardizing button styling in the settings panel (ref 26567)
• For our early beta testers of the KPI Tile with AI Insights we added a disclaimer to AI generated text used in the KPI tile. This disclaimer reads: "AI-generated content may be present. Check for accuracy. Here's a link to our AI usage details."
• We resolved a small visual bug where filter and breakdown selections in Waterfall charts were not staying in sync within the Define panel. Selections now consistently reflect the chart's configuration (ref 26455)
• Fixed an issue where SQL Explorer would render multiple spaces in a row as one space (ref 26654)
• We improved the error message shown when passing 0 as num_features in filter selection for One AI Machine Learning dimensionality reduction. The previous message was not helpful for troubleshooting. The new message is clearer: RuntimeError("0 is an invalid number of features for filter selection"). (ref ON-4083)
• In One AI Assistant Chat and Analyze modes, H1 and H2 headers are no longer included. These headers were visually oversized relative to the rest of the response. H3 or H4 is now the largest allowed header, improving readability and consistency. (ref ON-4177)
• Fixed an issue where using the Ctrl+Enter shortcut in SQL Explorer would override the Limit value configured in the SQL Explorer page. (ref 26026)
• We fixed a possible rounding error for Average, Median, and Percentile metrics, where the number of decimals rounded was based on the underlying column in the Processing Script, rather than the Data Type of the metric. This meant that if one took the average of three integers (say 1, 1, and 2), into a metric with two decimal places, then the result would still be 1 because of the datatype in the Processing Script, instead of 1.67, as per the datatype in the metric. (ref 19930)