Summary: Notion 3.2 arrived on January 20, 2026 with mobile AI Notes, the full Notion Agent on phones, and user-selected GPT-5.2, Claude Opus 4.5, or Gemini 3 while preserving context. Enterprise plans add AI usage analytics, while MCP audit logging and multi-database queries address governance. The release moves AI from a premium feature toward measurable daily operations and mobile agent workflows.
Notion opened its 2026 release calendar on January 20 with version 3.2 — the first release round of the year — under a theme the company describes as making Notion fit your world better. The headline push puts AI in your pocket: AI Notes turns a single tap into a recording that keeps transcribing, and the complete Notion Agent that lived on desktop now runs on phones. The change with bigger stakes is model choice — GPT-5.2, Claude Opus 4.5, and Gemini 3 are all on the menu, and the context you already have survives every switch.
Seen from a product team’s seat, no single feature carries the meaning of this update; what it confirms is a trend: AI has crossed over from premium add-on to table stakes in productivity tools, and “which foundation model am I talking to” has turned from an architecture decision into a setting users flip on a screen.
AI Notes and the Full Agent, Now in Your Pocket
On a phone, AI Notes begins transcribing after one tap and carries on even when you jump to another app or the screen locks; seconds after the meeting wraps, clear summaries, action items, and shareable docs are ready. The feature goes after the most practical headache of capturing meetings on a phone: recordings that cut out the instant the device drops out of the foreground. This time, background recording is written into the spec rather than wished for. With those instant post-meeting outputs on top, the meeting-notes workflow can drop both the separate recording app and the manual upload step.
The Notion Agent lands on mobile carrying everything the desktop version can do: forms get built, databases get created, company knowledge gets searched, and to-do items are worked through in the background while you turn to other things. An agent growing from desktop tool into an assistant you keep in your pocket is the loudest signal of this release — mobile agent capability is plainly the ground Notion means to fight on through 2026.
Switch Models Without Losing the Thread
With 3.2, picking GPT-5.2, Claude Opus 4.5, or Gemini 3 is directly in the user’s hands; anyone who would rather not decide can leave it to Auto to match a model to the task. Notion puts it bluntly: “Models will continue to change, but your memory in Notion doesn’t.”
That line lands on the core design problem of every multi-model product. Models themselves are interchangeable commodities; the asset is the context and memory a user builds up. Once a model switch stops wiping context, choosing a model comes down from a one-time technical wager to a preference toggle. Users get the convenience — and the platform gets the contest among three model vendors wired into its own product.
Enterprise Gets AI Usage Analytics
New on Enterprise plans is AI usage analytics, reached at Settings → Analytics → AI. From that panel, admins can tell who leans on AI hardest, which features genuinely pull people in, and where the value actually lands.
This is the familiar script of an AI feature entering the enterprise procurement cycle: capability first, then the demand for quantifiable proof of use. IT and operations teams that have to defend AI spend in front of leadership will keep meeting dashboards like this — until they quietly harden into a baseline item in vendor selection.
People Directory, Jira Sync, and Multi-Inbox Mail
People Directory gives everyone in the workspace a record in a database that comes pre-built and stays customizable, and it can import from HR systems; search “People” or go through Settings → Members → View People Directory to open it. Dev work flows into Notion through the Jira integration, where properties such as quarterly goals, customer feedback, and revenue impact can be layered on and merged into combined dashboards; Enterprise plans go further and edit Jira fields straight from Notion, with changes syncing back automatically. Notion Mail, meanwhile, handles multiple email addresses under one account and moves between personal and work inboxes.
Speed Gains and New Ground for MCP
The speed story: pages in the Windows app open 27 percent faster, the Mac desktop app gains up to 11 percent, and pages overall finish loading in roughly one-third less time. Enterprise MCP picked up upgrades as well — audit logging and multi-database queries have shipped, with controls over external AI tool connections promised next. Notion also looked back at 2025 as a year of 90-plus shipped features, counting Agents, AI notes, enterprise search, and offline mode among them.
Developers should read the MCP audit logs and external-connection controls as Notion engaging seriously with a governance question: enterprise data being exposed to third-party AI tools. The audit log matters most on its own terms, giving administrators a trail of what agents actually queried — the precondition for letting agent features loose on real workspaces. For teams, the release moves AI out of novelty and into daily operations, and the next battleground is clearly mobile, together with enterprise value you can measure.