This edition covers releases 2026.38 and 2026.40. The month's headline is a big one: Unique AI Conduct is now generally available. Alongside it come a clearer, more capable chat experience, a more productive Unique AI Table, and a fresh wave of AI models.
Unique AI Conduct Is Now Generally Available
Conduct, our agentic workspace for multi-step, end-to-end tasks, is now generally available, and the GA version is faster, more capable, and more trustworthy than the Beta: answers arrive several seconds sooner because the agent warms up as soon as you open a chat; spaces handle up to 400 turns per task, with time spent waiting for your input no longer counted; tool calls stream live and the agent pauses for your approval before acting in connected apps; and citation-aware grounding checks raise far fewer false warnings and remove footnotes without a real source.
A Clearer, More Capable Chat Experience
Unique AI Chat has a fresher look and you can now see the status of your chats at a glance.
Results from Code Execution (charts, files, and interactive HTML) now always appear as cards you can expand into the side panel.
Interactive HTML pages can do more: you can drop files onto a page to send them into the chat. HTML previews are now private to the person who generated them.
Better Document Translation
Translated PowerPoint and Word documents improved significantly and admins can also connect a translation memory, so recurring phrases match your previously approved wording.
A More Productive Unique AI Table
Bulk actions are now on by default, and they handle much larger selections: assign, unassign, or update review status on up to 50,000 rows at once. Besides, assignees on verified rows can be changed without unlocking their content.
Sheets now travel more easily between spaces: the Excel export carries each sheet's and row's tags, and those tags come back in when the file is imported elsewhere. Full Report Excel exports run on the server with a progress indicator for large tables.
Expanded Access to the Latest AI Models
Your model options keep growing. GPT-6 Astra, Luna, and Sol (via Azure), Gemini 3.8 Flash, Grok 4.6 and 4.7, and DeepSeek V4.1 Flash are all now available, and Conduct spaces can run on Claude Sonnet 5 and the new GPT-6 models.
For Platform Administrators
Admins get several new tools for governance and data quality. Microsoft Purview-protected Office files can now be set up for decryption during ingestion, so sensitivity-labelled documents can join your Knowledge Base. File metadata can be embedded with each chunk, helping search find the right documents, not just filter them. Admins can also enable and sign in to connectors such as Outlook and Salesforce directly from dashboards in chat.
Retention now goes further: once a file is removed, its name, description, and other identifying details go with it, and Conduct data is covered by the same tenant retention window. One heads-up: Code Execution is now switched on through a company-level setting, so if your team uses it, check with your customer success contact that it's enabled for your company.
What's Coming Next
A few features already in the pipeline are worth watching. Scheduled Tasks will let you edit, pause, and resume recurring tasks right from the answers they produce. In the Agentic Table, Edit with AI will help you refine an answer cell in a quick side chat, and Answer Library sharing will let owners grant edit and manage rights. A new document link graph will map the files a document links to, and Conduct will soon show source checks on the files it generates.
As always, these updates roll out automatically, with nothing for your team to install. If you'd like a walkthrough of Conduct or any of these features, or want to enable one that isn't yet active in your workspace, reach out to your customer success contact.
What's New in Unique AI in October 2026
This edition covers releases 2026.38 and 2026.40. The month's headline is a big one: Unique AI Conduct is now generally available. Alongside it come a clearer, more capable chat experience, a more productive Unique AI Table, and a fresh wave of AI models.
Unique AI Conduct Is Now Generally Available
Conduct, our agentic workspace for multi-step, end-to-end tasks, is now generally available, and the GA version is faster, more capable, and more trustworthy than the Beta: answers arrive several seconds sooner because the agent warms up as soon as you open a chat; spaces handle up to 400 turns per task, with time spent waiting for your input no longer counted; tool calls stream live and the agent pauses for your approval before acting in connected apps; and citation-aware grounding checks raise far fewer false warnings and remove footnotes without a real source.
A Clearer, More Capable Chat Experience
Unique AI Chat has a fresher look and you can now see the status of your chats at a glance.
Results from Code Execution (charts, files, and interactive HTML) now always appear as cards you can expand into the side panel.
Interactive HTML pages can do more: you can drop files onto a page to send them into the chat. HTML previews are now private to the person who generated them.
Better Document Translation
Translated PowerPoint and Word documents improved significantly and admins can also connect a translation memory, so recurring phrases match your previously approved wording.
A More Productive Unique AI Table
Bulk actions are now on by default, and they handle much larger selections: assign, unassign, or update review status on up to 50,000 rows at once. Besides, assignees on verified rows can be changed without unlocking their content.
Sheets now travel more easily between spaces: the Excel export carries each sheet's and row's tags, and those tags come back in when the file is imported elsewhere. Full Report Excel exports run on the server with a progress indicator for large tables.
Expanded Access to the Latest AI Models
Your model options keep growing. GPT-6 Astra, Luna, and Sol (via Azure), Gemini 3.8 Flash, Grok 4.6 and 4.7, and DeepSeek V4.1 Flash are all now available, and Conduct spaces can run on Claude Sonnet 5 and the new GPT-6 models.
For Platform Administrators
Admins get several new tools for governance and data quality. Microsoft Purview-protected Office files can now be set up for decryption during ingestion, so sensitivity-labelled documents can join your Knowledge Base. File metadata can be embedded with each chunk, helping search find the right documents, not just filter them. Admins can also enable and sign in to connectors such as Outlook and Salesforce directly from dashboards in chat.
Retention now goes further: once a file is removed, its name, description, and other identifying details go with it, and Conduct data is covered by the same tenant retention window. One heads-up: Code Execution is now switched on through a company-level setting, so if your team uses it, check with your customer success contact that it's enabled for your company.
What's Coming Next
A few features already in the pipeline are worth watching. Scheduled Tasks will let you edit, pause, and resume recurring tasks right from the answers they produce. In the Agentic Table, Edit with AI will help you refine an answer cell in a quick side chat, and Answer Library sharing will let owners grant edit and manage rights. A new document link graph will map the files a document links to, and Conduct will soon show source checks on the files it generates.
As always, these updates roll out automatically, with nothing for your team to install. If you'd like a walkthrough of Conduct or any of these features, or want to enable one that isn't yet active in your workspace, reach out to your customer success contact.