2026 July 22 - Release Notes

Welcome to the latest One Model product release update. This article provides an overview of the  product innovations and improvements to be delivered on 22 July 2026.

User Experience Improvements

One AI Assistant permissions simplified

• We've simplified how One AI Assistant permissions work so that each one stands on its own: grant any of Visualizations, Answers, Chat, or Storyboard Navigation and the Assistant appears for that user with just that capability. (Ref 28952)
• We've also removed the previous permission prerequisites for Insights and Analyze. Nothing changes about where these features appear - they still only surface on content the user already has permission to view (such as a visualization, Answer, or Storyboard tile) - and existing roles are unaffected. (Refs 28938, 27945)

AI-generated content is now clearly labelled

• AI-generated text responses in One AI Assistant and on Storyboard tiles now carry a clear “AI-generated” label next to the content. This covers Chat responses, the Answers explanation, Analyze responses, and Insights text. Charts and KPI numbers are not labelled, since they're built directly from your data. This supports EU AI Act Article 50 transparency requirements. (Ref 27884)

Data Integration Innovations

• We have enhanced our JSON file processing. Every JSON record processed through our data converter now gets a consistent, unique row identifier, regardless of source system. This makes it easier to join and cross-reference data across tables, giving your analytics and data teams more reliable lineage when building custom reports or validating data integrity. This new GUID-based Row ID generation is for JSON files only. (Ref 28970)
• We fixed an issue in the Data Loading Service where a retried message could create a duplicate DataLoad and DataLoadFile for the same uploaded file. Upload, antivirus scan, and SFTP extraction now check for an existing DataLoadFile before creating a new one, so retries continue against the original DataLoad instead of creating an orphaned duplicate. (Ref 28037)
• We improved the resilience of our database loading service against Redshift serialization errors encountered during data loads, adding retry handling so transient conflicts are less likely to require a manual retry. (Ref 27964)
• Deleting a data source or input (Delimited, JSON, XML, or Compressed Folder) no longer prompts you to separately choose whether to remove its staging tables from the data warehouse. Deletion now always cleans up both the source/input metadata and its staging tables in Redshift, so no orphaned tables are left behind. You'll still see a progress message directing you to the Data Loads page to monitor cleanup, and you can retry the delete if the item reappears. (Ref 27385)
• The API Runs page now displays any messages generated during an API run. A Messages field in the run detail panel shows the message count, and clicking it opens the full list. (Ref 26343)

Data Processing Innovations

• We added support for the current_date keyword in Processing Scripts, as an alternative to GETDATE() that returns a date instead of a datetime. This can reduce Processing Script run times where there are multiple Data Loads on the same day, as objects with current_date update only once per day (or when data changes), instead of reprocessing them on every Data Load as they would with GETDATE(). (Ref 28996, 27117)
• We’ve rolled out performance improvements to our standard Average Headcount calculation, and similar queries. Depending on your context, uncached queries can now run over 50% faster. This had been gradually being rolled out across customers, but is no officially live for everyone. (Ref 18112)
• You can once again create table-based metrics that combine a column from a splice-joined-to-time table with a Time operand — a restriction introduced during Average Headcount validation has been lifted. (Ref 29040)
• The new Average Headcount calculation is now the standard method for headcount metrics; the legacy feature flag has been retired. The automatic-rollback safeguard remains in place. (Ref 18112)

Minor Improvements & Bugs Squashed

Insights & Analyze fixes

• We fixed another cause of the “Exceeded retry limit” error in Auto Insights on Storyboard Tiles. This message can result from several distinct underlying data conditions. Insights now generate reliably for the affected tiles. (Ref 28988)
• We fixed an issue where opening a geo map tile via Analyze in the One AI Assistant showed a blank map. The map now renders correctly. (Ref 28332)
• We fixed an issue where asking Analyze to analyze a chart that returned no data would leave the assistant stuck in an “Analysing” state indefinitely. (Ref 28129)
• Insights and Analyze are now disabled whenever a chart or table shows the “Data limit reached” message, since the underlying data is incomplete. Previously these tools could still run against truncated data, producing inaccurate results. Refine or export the query to return the full dataset to use Insights or Analyze again. (Ref 27774)

Other minor improvements and fixes

• When you pin tiles from the Insight Library to your Home Page, the pop-up now closes automatically and a confirmation banner appears. If you're already viewing your Home Page storyboard, it refreshes so newly pinned insights show up right away. (Ref 28879)
• We fixed a bug where Ghost Buttons displayed an inherited background fill from the Standard Button styling instead of remaining transparent. Ghost Buttons now render with no fill, restoring the visual distinction from standard buttons. We also fixed an issue where the tab's overflow (three-dot) menu displayed a white box instead of matching the tab's color. The overflow menu now correctly inherits the tab's color. (Ref 28513,  28721)