# Gemini in meet, drive, and the side panel
Gemini takes notes in your Meet call, then helps you find and draft from the related Drive doc without you ever switching tabs. That cross-app continuity is the real story of Gemini in Workspace: the same model follows you from a live meeting into the document where the work actually happens. This lesson walks one concrete workflow end to end, then shows where the automation hooks are when clicking around stops scaling.
The Gemini side panel is the assistant that opens on the right edge of Docs, Gmail, Sheets, Slides, and Drive. It is not a separate chatbot. It is context-aware: it knows which file you have open, and it can reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → across your other Workspace content when you ask.
Two things make it different from the standalone Gemini app:
That grounding is the key word. Instead of the model guessing from training data, it retrieves from *your* documents and links back to them. You stay in control of the source.
Availability depends on your Workspace edition and admin settings. Check the Workspace support hub for Gemini for what your domain has enabled.
In Google Meet, "Take notes for me" captures the meeting as it happens. It is a separate feature from the live captions and from the recording: notes is a structured summary, captured by Gemini, that lands in a Google Doc after the call.
What you get when the meeting ends:
The practical upgrade over old-style transcripts: action items are extracted, not just transcribed. You walk out with "Priya to confirm vendor SLA by Friday" instead of forty paragraphs you will never reread.
A few things worth knowing so you trust the output:
Here is the move that the side panel is actually built for. You just finished a planning call. The notes Doc exists. Now you want to act on it without rebuilding context.
Step 1: Meet generates the notes. Call ends. A Doc like *"Q3 Launch Planning, Notes by Gemini"* appears in your Drive, with action items at the top.
Step 2: Open the notes Doc and ask the side panel to connect it to prior work. In the Doc, open Gemini and prompt:
> Compare these action items to the "Q3 Launch Plan" doc in my Drive. What's new, and what conflicts with what we already committed to?
Gemini reads the open Doc, finds the older plan in Drive by grounding, and returns a diff with links to both files. You did not search Drive manually. You did not paste anything.
Step 3: Draft the follow-up from the side panel. Still in the Doc:
> Draft a follow-up email to the launch team summarizing the three new action items and flagging the date conflict. Keep it under 150 words.
Move to Gmail, open the side panel there, and ask it to "help me write" using that summary. The draft inherits the meeting context because you are working off the same grounded sources.
The whole loop, Meet to Drive to Gmail, never leaves the side panel. That continuity is the feature. The model is the same; the *context* travels with you.
When you paste a transcript into a chatbot, the model only knows what you pasted. The side panel's Drive grounding means it can pull the *older* plan you forgot existed, cite it, and reason across both. You get fewer hallucinated "facts" because the answer is anchored to retrievable files, and you get an audit trail because every claim links to a source.
This is retrieval over your own corpus, done for you, without you standing up a RAG pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition →.
The side panel is great for one meeting. It does not scale to "do this for every sales call and log it to a tracker." That is where Apps Script and the Gemini APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition → come in.
Apps Script is Google's built-in JavaScript automation layer for Workspace. You can call Gemini from it to process documents on a schedule or a trigger. Here is a clean example: a function that reads a meeting-notes Doc, asks Gemini to extract action items as structured JSON, and returns them so you could append them to a tracking Sheet.
function extractActionItems(docId) {
const text = DocumentApp.openById(docId).getBody().getText();
const apiKey = PropertiesService.getScriptProperties().getProperty('GEMINI_API_KEY');
const url = 'https://generativelanguage.googleapis.com/v1beta/' +
'models/gemini-2.5-flash:generateContent?key=' + apiKey;
const prompt = 'Extract action items from these meeting notes. ' +
'Return JSON: [{"owner": string, "task": string, "due": string}].\n\n' + text;
const response = UrlFetchApp.fetch(url, {
method: 'post',
contentType: 'application/json',
payload: JSON.stringify({
contents: [{ parts: [{ text: prompt }] }],
generationConfig: { responseMimeType: 'application/json' }
})
});
const result = JSON.parse(response.getContentText());
return JSON.parse(result.candidates[0].content.parts[0].text);
}Two senior-engineer notes on this:
responseSchema for production.You would trigger extractActionItems from a Drive change event or a daily run, then write the JSON rows to a Sheet. Now every meeting note flows into one tracker automatically. Get an APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition → key and test prompts first in Google AI Studio, and read the APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition → reference at ai.google.dev.
If you find yourself typing the same long instruction into the side panel every meeting, build a Gem instead. A Gem is a saved, custom version of Gemini with fixed instructions. Create a "Meeting Follow-Up" Gem with a system prompt like "Always extract action items with owners and due dates, flag conflicts with existing plans, and draft a sub-150-word email." Then you invoke behavior, not paragraphs.
Gems live in the Gemini app and, increasingly, surface across Workspace. They are the no-code complement to the Apps Script route: Gems for personal reuse, Apps Script and the APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition → for systematic automation.
Knowledge check
1. What does it mean that the Gemini side panel is 'context-aware'?
2. Why is 'grounding' described as the key concept behind the side panel's answers?
3. What is the practical advantage of Meet's 'Take notes for me' over an old-style transcript?
4. Select ALL statements that correctly distinguish the Gemini side panel from the standalone Gemini app.
Select all the correct answers.
5. Select ALL that accurately describe what you receive after a Meet call using 'Take notes for me'.
Select all the correct answers.
The Drive side panel turns your file store into something you can interrogate. Open Drive, open Gemini, and ask:
> Summarize the three most recent docs in the "Vendor Contracts" folder and tell me which ones expire this quarter.
It grounds in those files and answers with links. This is the same retrieval capability that powers the Meet-to-Drive step above, just invoked directly.
Where it pays off:
Limits to keep honest: Gemini grounds in content you have permission to access, and quality depends on your files actually saying what you think they say. Stale or contradictory docs produce stale or contradictory answers. The model is only as good as your Drive hygiene.
The side panel and Apps Script cover individual and team automation. When you need org-wide deployment, governance, and your own dataown dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.View full definition →, the same Gemini models are available through Vertex AI on Google Cloud, where you get enterprise controls, grounding on your private data stores, and the infrastructure to run agents at scale. See cloud.google.com/vertex-ai. The mental model: Workspace side panel for in-the-flow assistance, the Gemini APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition → plus Apps Script for Workspace automation, Vertex AI when you are building a governed product.
You do not jump straight to Vertex for a meeting-notes workflow. Start where the work is. Graduate when the requirements (security, scale, custom data) demand it.
responseMimeType, and reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it. for the Flash tier on high-volume extraction.These actions are compiled in the role's Playbook.