Tired of digging through folders for project information? This playbook shows you how to build a custom research agent inside Gemini Enterprise that searches your Google Drive — and optionally Gmail and Calendar — to answer complex, project-specific questions.
Ask things like:
“Find all notes and emails about Client X’s marketing strategy and summarize the key findings.”
The agent searches your connected data, summarizes what matters, and even cites its sources — all inside a secure, permission-aware environment.
What You’ll Need
Before starting, make sure you or your Workspace admin has:
- A Google Workspace account with Gemini Enterprise enabled
- Permission to connect Drive, Gmail, and Calendar data
- Access to a shared drive or project folder that holds your documents
- (Optional) Admin rights to enable APIs if you want the agent to take actions (like emailing results or booking meetings)
Step 1 — Connect Google Drive
- Go to Google Cloud Console → Gemini Enterprise → Data Stores → Create Data Store.
- Choose Google Drive as your source.
- Select your scope:
- All of Drive (entire domain)
- Specific shared drive(s)
- Specific folder(s) — paste the folder ID from Drive.
- Click Create.
Pro Tip: Keep project docs in a domain-owned shared drive — Gemini can only search data your Workspace owns and that you personally have access to.
Step 2 — (Optionally) Add Gmail and Calendar
To include context from conversations or meetings:
- Gmail: In Data Stores → Create Data Store → Gmail
- Calendar: In Data Stores → Create Data Store → Google Calendar
These enable questions like:
“Include recent emails or meetings about Client X’s strategy.”
Step 3 — Create a Gemini Enterprise App
- Go to Gemini Enterprise → Apps → Create App.
- Name it something like Research Assistant.
- Attach your data stores: Drive, Gmail, and/or Calendar.
- Save the app — this becomes the foundation for your agent.
Step 4 — Build Your Research Agent
You’ll use the Agent Designer (no-code interface) to define the agent’s behavior.
- In your new app, go to Agents → Create Agent.
- Fill in the details:
- Name: “Client Research Assistant”
- Goal: “Answer project questions using Drive (+ Gmail/Calendar if enabled).”
- Instructions: Paste the template below.
- Data Sources & Tools: Add your Drive (and optional Gmail/Calendar).
- Starter Prompts: e.g. “Summarize all notes on Client X’s marketing strategy from Q2.”
- Test in the preview panel.
- Click Save.
Agent Instruction Template
You are a research assistant for our {team}.
Use the attached data sources (Google Drive, and if available Gmail and Calendar)
to answer questions with citations (file titles + links where possible).
Prioritize:
1. The most recent and most relevant content first.
2. Executive summaries before details, unless asked otherwise.
When asked for “all notes/emails/meetings about {Client/Project},” do:
- Search Drive for project docs, briefs, and notes.
- If available, include relevant Gmail threads and Calendar meetings.
- Aggregate themes, decisions, next steps, and responsible owners.
- Provide a bullet summary followed by a short “What This Means” section.
- If you don’t have enough data, say so and suggest where to look.
Step 5 — (Advanced) Enable Actions
Want your assistant to email summaries or book review meetings?
- Ask your admin to enable:
- Gmail API
- Google Calendar API
- Google People API
- In Google Cloud Console → OAuth Consent Screen, set up an Internal App.
- Redirect URI:
https://vertexaisearch.cloud.google.com/oauth-redirect
- Redirect URI:
- Copy the Client ID and Client Secret.
- In your Gemini App, go to Actions → Add Action → Gmail or Calendar,
paste the credentials, and authorize.
Now your agent can send summaries or create follow-up meetings automatically.
Step 6 — Test It
Try queries like:
“Find all documents about Client X’s marketing strategy since July 1 and summarize key findings.”
“Include relevant emails from the last 60 days. List decisions, owners, and next steps.”
“What were the main meeting outcomes for Client X in Q3?”
Gemini will return summarized results, grouped by source and linked back to the original files — respecting your Workspace permissions.
Step 7 — Refine and Monitor
- Keep content in shared drives (Gemini can’t search private My Drive items).
- Use clear file names and metadata so results cluster well.
- Use Prompt Chips (admin feature) to standardize common questions.
- Check Analytics under your app to see what users are asking and fine-tune the agent’s instructions.
Troubleshooting
| Issue | Cause | Fix |
|---|---|---|
| File missing from results | Not in a domain-owned drive or too large/unscanned | Move to a shared drive; check OCR/file size limits |
| 403 Permission error | Using a service account | Use a Workspace user account |
| Agent can’t email or book meetings | APIs/OAuth not set up | Enable Gmail/Calendar APIs and attach action credentials |
| Slow or partial results | Large or unindexed data store | Limit scope to project folders or use filters |
What You’ll End Up With
- A personal AI research assistant grounded in your Drive, Gmail, and Calendar
- Secure, permission-aware summaries with citations
- Optional ability to email findings or schedule meetings automatically
- A foundation you can clone and adapt for any client, project, or department
Next Steps
- Train your team: Encourage consistent naming and file organization.
- Add prompt chips: e.g., “Summarize all client notes from this quarter.”
- Expand to other connectors: Add Jira, Confluence, or CRM data sources next.