The raw material
Modern AI depends on very large collections of information: text, images, code, websites, documents, and other examples. The quality of the data shapes the quality of the answer.
Company field guide
Your AI
Guide
A practical starting point for using AI at work: what AI is, what tools are available through Google Workspace, when to use Gemini or NotebookLM, and how to write prompts that get useful results.
Start here
What is AI?
AI is software that can recognize patterns, make predictions, and generate new text, images, summaries, plans, or analysis based on the information it has been trained on and the information you give it.
The pattern finder
An algorithm is the set of instructions the system uses to learn from data. Training can take months, and a model may not know about recent events unless the tool can search the web or you provide current sources.
The trained system
Large language models work with words. Large visual models work with images. Multimodal models can work across text, images, files, and other inputs.
The tool you use
Gemini, ChatGPT, Claude, and NotebookLM are interfaces built around models. The interface matters because it controls memory, file uploads, settings, source handling, and how much context the tool can use at once.
AI is useful, but it still needs judgment.
AI can draft, summarize, compare, brainstorm, classify, and turn messy material into a usable first pass. It can also be confidently wrong, outdated, too generic, or too polished. Treat it as a capable assistant that needs clear direction, good source material, and human review.
Access
Why use Google?
Your company Google account gives you access to the Google Workspace suite, including approved Gemini features for private text assistance, image generation, image analysis, and document-based work. Start from your work email so files, permissions, and collaboration stay inside the company environment.
Creating, drafting, and exploring
- Brainstorming ideas, outlines, titles, emails, agendas, and talking points.
- Rewriting text for clarity, tone, length, or audience.
- Image generation, image analysis, quick explanations, and early creative thinking.
- Turning a rough thought into a first draft you can edit.
Studying sources and synthesizing research
- Working from a defined set of documents, reports, meeting notes, PDFs, or web sources.
- Summarizing source material with citations back to the files you provided.
- Comparing documents, pulling themes, creating FAQs, or preparing study guides.
- Research synthesis when accuracy depends on the uploaded sources.
Good practice
Use Gemini when you need a creative or drafting partner. Use NotebookLM when the answer must come from specific sources.
Privacy
Use the work version for work.
Company AI work should happen inside the company Google Workspace environment. That keeps files, access, permissions, and account controls tied to the business instead of a personal account.
It gives the company more control.
Using Google Workspace Business helps keep work activity inside a managed environment. The company can manage users, permissions, file sharing, security settings, and access if someone changes roles or leaves.
Use judgment before uploading.
Do not add confidential, personnel, financial, legal, client-sensitive, or private personal information unless the tool and the use are approved for that material. Remove details that are not needed for the task.
Keep personal use separate.
The work version is for work. Do not use the company AI environment for personal conversations, private life planning, health questions, personal finances, or anything you would not want tied to your work account.
Compare tools with safe material.
When the material is non-sensitive, it is useful to test the same prompt in multiple models. Compare clarity, accuracy, structure, creativity, and source handling before deciding which tool fits the job.
Access
Do not run agents or bots on local machines.
AI assistants such as Gemini or NotebookLM are user-facing tools. Agents and bots are different: they can take actions, run code, browse files, call services, or automate workflows. That makes them powerful, but it also makes local access risky.
Keep autonomous tools out of the local environment.
Do not install or run AI agents, bots, coding agents, automation scripts, or autonomous browser tools on a work laptop or desktop unless the company has explicitly approved that setup. Local machines may contain client files, credentials, browser sessions, synced drives, and internal communications that an agent should not be able to reach by default.
Use a VPS or other isolated environment.
If an agent or bot needs to run, run it through a controlled VPS, cloud workspace, or sandboxed environment with only the minimum files, credentials, and permissions needed for the task. Treat that environment as disposable, monitorable, and separate from personal or company laptops.
How to ask
Prompting guide
Good prompts are simple, specific, and broken into steps. You do not need fancy language. You need clear instructions.
- Say what you want. Name the task plainly: summarize this, rewrite this, make an outline, compare these options, draft an email, or create a checklist.
- Give context. Tell the tool who the audience is, what the work is for, what background matters, and what source material it should use.
- Set the role or mode. Ask it to act like an editor, project manager, planner, researcher, trainer, or reviewer. Use “correct mode” for the work you need.
- Break the work into steps. Ask for one step at a time when the task is complicated. For example: first summarize, then identify gaps, then draft the final language.
- Specify the output. Tell it the format: bullets, memo, table, email, agenda, slide outline, checklist, short paragraph, or step-by-step instructions.
- Ask it to check itself. Have it flag assumptions, missing information, weak spots, or claims that need a source.
Copy-and-paste starter prompt:
I need help with [task].
Audience: [who will read or use this].
Context: [what this is for and what matters].
Use these sources: [paste text or upload files].
Please do this step by step:
1. Summarize the key points.
2. Identify gaps, risks, or unclear areas.
3. Draft the final output in [format].
Tone: [plain, professional, friendly, concise].
Before finalizing, list any assumptions or facts I should verify.
AI does not remember everything forever.
Some tools have memory, some do not, and every chat has a context limit. If a conversation gets long, the tool may lose track of earlier details or start mixing instructions.
Start fresh when the topic changes.
Use a new chat for a new task, a new client, a new project phase, or a new set of instructions. A clean chat usually gives cleaner results.
Bring the important context with you.
If an old chat contains useful work, export or download it when the tool allows, then upload or paste the relevant parts into the new chat. Add a short summary of what matters most.
Save repeatable instructions.
For recurring work, create a standard instruction set or Gemini Gem. Include the role, audience, tone, do-and-don’t rules, source preferences, and output format. This helps the tool behave consistently across tasks.
Team habits
How to get better results over time
The team will get more value from AI by improving inputs, testing tools, and building reusable instructions.
- Clean the data set. Keep project files, notes, images, reports, and source documents organized so AI tools can work from better inputs.
- Test the tools with non-sensitive material. Try the same safe prompt in Gemini, NotebookLM, or another approved model. Compare the results. Notice which tool is better for drafting, analysis, source work, or synthesis.
- Draft standard prompts. Create reusable prompts for common needs: meeting summaries, project research, proposal outlines, slide drafts, document review, and client-ready language.
- Review before sharing. AI output should be checked by a person before it becomes a deliverable, email, proposal, or public-facing document.
Bottom line
AI is strongest when people control the inputs, define the task, and review the output. The goal is faster first drafts, clearer synthesis, and better organized thinking.