The Brief

Google has begun rolling out deep Gemini integration across Docs, Sheets, Slides, and Drive, enabling AI-assisted drafting that draws on a user’s emails, chats, and files. A new “Fill with Gemini” feature in Sheets auto-populates data at what Google claims is nine times the speed of manual entry, while Drive gains cross-document search and Q&A capabilities. The features are available in beta to AI Pro and Ultra subscribers, with Drive features initially limited to the US.

The Report

Google announced a sweeping expansion of Gemini’s role across its Workspace productivity suite on March 10, giving the AI model direct access to users’ Gmail, Google Chat, Drive files, and Calendar data to generate documents, spreadsheets, and presentations from natural-language prompts.

The centrepiece is a reworked creation flow. In Docs, a new “Help me create” prompt bar lets users describe what they need — a campaign plan, a newsletter, a project brief — and Gemini assembles a formatted first draft by pulling relevant information from across their Workspace data and the web. Companion features allow users to unify writing style across a document or transfer the formatting of one file to another. In Sheets, users can now build entire spreadsheets through conversation, with Gemini orchestrating data gathering and complex calculations. The “Fill with Gemini” function auto-populates cells by analysing adjacent data, column headers, and — where needed — live web results.

Google cited a 95-participant internal study in which Fill with Gemini completed 100-cell data tasks nine times faster than manual entry. On the public SpreadsheetBench benchmark, Gemini scored 70.48% — marginally below the human baseline of 71.33% but ahead of Kingsoft’s Qingqiu Agent at 69.96%. Google described the performance as “near-human level.”

Drive received what one analyst called the “sleeper” update: AI-generated overviews now appear at the top of search results with citations, and a new “Ask Gemini in Drive” feature allows cross-document questioning spanning files, emails, and calendar entries. Mike Leone, practice director at Omdia, said the changes could shift Drive “from passive file storage into something closer to an active knowledge base.”

Industry response was measured. Julie Geller of Info-Tech Research Group called the updates “incremental improvements more than revolutionary features, but they address real workflow gaps.” Amy Machado of IDC noted that accuracy remains “a top challenge for genAI” and that user trust is “the final boss.” She assessed Google’s announcement as “on par with what Microsoft offers” — Microsoft having unveiled its own Copilot Cowork agent system one day earlier.

The rollout arrives as Google Workspace serves three billion monthly active users and approximately eleven million paying business customers. Google stated that company data used to ground Gemini responses is not used to train its global models, though a separate California class-action lawsuit alleges the company has made opting out of consumer-facing Gemini access in Gmail intentionally difficult. The features are available in beta to Google AI Pro ($20/month) and Ultra subscribers, with Docs, Sheets, and Slides available in English globally and Drive features initially restricted to US customers.

The Angle

The timing tells its own story. Microsoft ships Copilot Cowork on a Sunday. Google ships Gemini-everywhere on a Monday. The product announcements are different in architecture — Microsoft leans toward autonomous agents that run for hours, Google toward assistants embedded inside existing documents — but the strategic intent is identical: make the AI layer so deeply woven into the productivity suite that switching costs become functionally infinite.

That is the actual product being shipped here. Not a spreadsheet feature. Not a drafting tool. The product is dependency — measured in how much of your organisational knowledge flows through a system that now actively reads, interprets, and acts on it. Google’s privacy assurance that enterprise data will not train its global models is doing considerable load-bearing work in that sentence, and the ongoing class-action over consumer Gmail opt-outs suggests the wall between enterprise protections and consumer defaults is thinner than the reassurance implies.

The analyst consensus — “solid but not groundbreaking” — is probably the accurate near-term read. AI-assisted document drafting has existed for two years. What has changed is not the capability but the surface area. Three billion monthly users are now one toggle away from an AI layer that reads their email, scans their files, and populates their spreadsheets with data it sourced from the web. The question nobody in the announcement addressed is what happens to the skill of populating a spreadsheet — or drafting a brief, or structuring an argument — when the default workflow is to describe what you want and let the system build it. Nine times faster is a speed metric. It is not an understanding metric. The gap between those two measurements is where the interesting consequences will emerge.