Gemini moves into Workspace

Google has folded its Gemini AI into the core of Workspace, turning the assistant from a separate chat toy into a feature that reaches into Docs, Sheets, Slides, Drive, Meet, Gmail and a video app called Vids. (blog.google) The change looks like this in practice: Gemini can generate a first draft for a document by pulling details from your Drive files and emails, then refine tone and format on demand inside Docs’ side panel. (blog.google) In Sheets, Google says Gemini can edit whole spreadsheets on request — not just suggest formulas but perform complex manipulations and analyses when you describe the goal. (blog.google) Vids, Google’s video creation tool, now exposes directable AI avatars and automatic soundtrack generation so teams can create on‑brand clips without a studio; the models behind this are Veo 3.1 for video and Lyria 3 for music. (workspace.google.com) (blog.google) Meet gets wider language support for its “Ask Gemini” feature: the assistant can now provide in‑meeting help and translated audio in multiple languages so participants hear a near real‑time translation of speech rather than only captions. (support.google.com) (pcmag.com) One unifying change is retrieval: Gemini can be pointed at particular data sources — specific files, emails, or apps like Chat — and then synthesize answers from them instead of relying only on a global model context. (blog.google) (workspaceupdates.googleblog.com) That retrieval behavior comes with permissions baked in. Conversation history and Gemini’s ability to reference prior chats are implemented with app‑specific scopes: a chat you have with Gemini inside Docs won’t automatically appear inside Sheets, and admins can control retention and deletion. (workspaceupdates.googleblog.com) (support.google.com) Google is gating some of the more powerful features to paid tiers: the March rollout names Google AI Pro and Ultra plans as the first recipients for several new capabilities. (blog.google) Why this matters for a software‑engineering student thinking about real systems: the product shows AI being designed as an orchestration layer rather than a single app. Gemini’s usefulness depends on three engineering pieces working together — fast retrieval from multiple stores, clear permission boundaries so data exposure is predictable, and consistent UI affordances across different apps so users can steer the model. (blog.google) (workspacegoogle.com) From a systems perspective those pieces map to concrete challenges: indexing and vectorizing documents for low‑latency retrieval, enforcing ACLs at query time, and designing side‑panel interactions that preserve conversational context while avoiding data leakage. (blog.google) The rollout is incremental and controlled: Google published admin controls and timelines, and began making conversation history and the new Vids capabilities available this spring to eligible accounts. (workspaceupdates.googleblog.com) (workspace.google.com) For developers building portfolio projects, the takeaway is practical: users expect AI to be able to fetch and act on real data while respecting access control and offering clear UI patterns for “undo” or deletion. The Workspace changes make those expectations explicit. (blog.google) Administrators who want to test or restrict these features can do so now — Google listed that admin controls for conversation history were available as of February 25, 2026. (workspaceupdates.googleblog.com)