Gemini inside Google Workspace

Gemini in Workspace is the version that sits inside Gmail, Docs, Sheets, Slides and Meet. It is worth treating as a distinct tool from the Gemini chat app, because the thing that makes it useful — access to your existing content — also sets its limits.

Where it earns its place

• Mail. Summarising a long thread before you join it, and drafting replies that already know what the thread said. This is the feature most people keep using after the novelty passes.
• Docs. Drafting into an existing document, restructuring a rough section, adjusting tone across a whole piece rather than a paragraph.
• Meet. Notes and summaries from a call you attended or missed. See AI for meetings for how much to trust these.
• Sheets. Explaining what a formula does, generating one from a description, and organising messy data — genuinely useful, and the area where checking the output matters most.

Where it disappoints

Expect the assistance to be shallower than the chat app on hard reasoning tasks. The in-document experience is optimised for quick, contextual help, not for extended back-and-forth. When a task needs real thinking, people generally end up in the standalone app with the content pasted in — which is a reasonable workflow, just not the one the integration promises.

The other common disappointment is retrieval. “It should know this, it’s in my Drive” runs into the reality that finding the right document among thousands is hard, and Gemini will sometimes confidently work from the wrong one, or from nothing at all. Naming the document you mean is far more reliable than hoping it infers which you meant.

The integration’s value is proximity, not intelligence. It saves you the copy-paste, and that is worth a lot on routine work and very little on difficult work.

Getting more out of it

The same prompting habits apply, with one addition: be explicit about which content you want used. “Summarise the Q3 budget doc” beats “summarise the budget doc” when several exist, and pointing it at a specific file beats both.

Give it the constraints you would give a colleague — audience, length, tone, what to leave out. The model is not short of capability on these tasks; it is short of context about what you actually want, and that is the part you control.

Before you use it on anything sensitive

Read Gemini’s data controls. The short version: enterprise content is not used for training and Gemini respects existing Workspace permissions — but it makes existing over-sharing much easier to stumble into, which is a different problem and a real one.