AI Tools, Prompting & Your Pro Workspace

Welcome. This module assumes you are curious, capable, and new to professional AI work. You do not need to know code. You do not need ten subscriptions. You need a clear mental model, a small toolset, and a repeatable way to turn an idea into reliable work.

Level: Beginner
Time: 75–100 minutes
Outcome: A working AI workspace, a reusable prompt template, and your first live project workflow.

What you will be able to do

By the end of this module, you will be able to:

• explain what generative AI is good at—and where it can confidently fail;
• choose an appropriate AI tool for writing, research, learning, files, visuals, or coding;
• write clear prompts using a simple professional structure;
• improve weak answers through an intentional review loop;
• set up an organized workspace that you can use for every future project;
• protect private information and verify important claims; and
• run a project in a live local environment and understand what is happening.

1. First, change how you think about AI

An AI assistant is not a magic answer machine. Think of it as a very fast collaborator with three unusual traits:

1. It has broad pattern knowledge.
2. It can produce and transform work quickly.
3. It does not automatically know your real goal, private context, or definition of “good.”

That third point is why professional AI users get better results. They do not merely ask questions. They brief, review, and direct the system.

A useful working model is:

The goal of this course is not to make you dependent on AI. It is to make you better at directing work, evaluating quality, and building systems with AI.

2. What AI is good at

Use AI when the task benefits from language, patterns, transformation, or iteration.

Strong use cases

• Understanding: explain a concept at your level, compare ideas, create examples, or quiz you.
• Writing: outline, draft, rewrite, summarize, change tone, or adapt content for a different audience.
• Research support: build a research plan, find sources, compare evidence, and identify unanswered questions.
• Analysis: organize messy notes, extract themes, classify information, or inspect a spreadsheet.
• Creation: brainstorm names, design a content plan, generate visual directions, or prototype an interface.
• Technical work: explain errors, plan a feature, write code, test a change, or document a system.
• Operations: turn a repeated process into a checklist, template, standard operating procedure, or automation plan.

Weak or risky use cases

• trusting a factual answer without checking its source;
• asking for medical, legal, or financial decisions without a qualified professional;
• uploading passwords, private client data, identification documents, or confidential files without approval;
• letting AI send, publish, purchase, delete, or change access without reviewing the exact action;
• using a polished answer as proof that the answer is correct.

Fluency is not accuracy. AI can sound certain when it is mistaken. Professional use always includes verification proportional to the risk.

3. Your essential AI tool stack

Do not collect tools. Build a small stack in which every tool has a job.

Start with these five roles

Which assistant should I choose?

Choose one primary assistant for your first month. The best one is the one you can access consistently and learn deeply.

• ChatGPT: a strong general workspace for conversations, files, web search, data analysis, images, and long-running Projects. Use a Project when the work will continue across several sessions.
• Claude: useful for careful writing, long documents, analysis, and self-contained Artifacts such as documents, diagrams, or small interactive tools.
• Gemini: useful when your work already lives in Google’s ecosystem; its research and workspace connections can reduce copying between tools.
• Perplexity: useful when the main task is web research and you want visible sources early in the process.
• NotebookLM: useful when the answer should come from a defined source collection rather than the open web.

Features and free-plan limits change. Never build your learning around a feature name. Build it around the role: collaborate, research, ground, organize, build.

The minimum-cost starter stack

You can begin without paying for multiple products:

1. One free general AI account.
2. Google Drive or a local folder for your files.
3. NotebookLM for source-grounded study when available.
4. VS Code for editing projects.
5. GitHub for version history and publishing public practice projects.

Upgrade only when a real limit repeatedly blocks useful work. “It looks more professional” is not a reason to subscribe. “I lose two hours every week because I hit this limit” may be.

4. Prompting: give a brief, not a wish

A prompt is the instruction and context you give an AI system. A professional prompt does not have to be long. It has to remove the ambiguity that matters.

Use the BRIEF method:

Weak prompt

Teach me AI tools.

This is not wrong, but the assistant has to guess your level, goals, time, preferred examples, and desired output.

Strong prompt

Background: I am a beginner who uses a Mac and wants to become confident with AI for research, writing, and building simple websites.
Result: Create a seven-day practice plan that takes 45 minutes per day.
Inputs: I currently use ChatGPT, Google Drive, Chrome, and VS Code.
Expectations: Use plain English. Introduce only one new concept per day. Include a real exercise and a definition of “done.” Do not assume I can code. Flag anything that requires a paid plan.
Format: A table with Day, Skill, Exercise, Prompt to Try, and Completion Check.

The second prompt gives the AI something it can evaluate its own response against.

Your reusable prompt template

Copy this into a note called Prompt Template:

Background:
I am...
This matters because...

Result:
I need you to produce...
The work will be used for...

Inputs:
Use these facts/files/examples...
If information is missing, identify it instead of inventing it.

Expectations:
Audience:
Tone:
Must include:
Must avoid:
Quality standard:

Format:
Return the answer as...

Before starting:
State your understanding in one sentence and list any assumption that could materially change the result.

You will not need every field every time. For a quick rewrite, one sentence may be enough. For important work, a structured brief saves time.

5. The five-turn professional workflow

Beginners often try to get a perfect answer in one prompt. Professionals use a loop.

Turn 1 — Align

Ask the AI to restate the goal and expose assumptions.

Restate my desired outcome in one sentence. Then list the three assumptions most likely to affect your answer.

Turn 2 — Plan

Ask for the approach before the full output when the task is complex.

Propose a short plan. For each step, explain what evidence or input you need. Do not create the final deliverable yet.

Turn 3 — Create

Approve or correct the plan, then request the first version.

Use the plan, with these changes: [changes]. Create version one. Mark any unverified claim with [VERIFY].

Turn 4 — Critique

Do not ask only “make it better.” Define the review lens.

Review this as a skeptical customer. Identify unclear claims, missing evidence, unnecessary jargon, and the three places most likely to lose trust.

Turn 5 — Revise and verify

Revise using the critique. Preserve the strongest parts. End with a verification checklist separating confirmed facts, assumptions, and decisions still needed.

This loop works for an email, research report, landing page, spreadsheet analysis, course lesson, or software feature.

6. How to talk to AI naturally

Structured prompting does not mean robotic prompting. Talk to the assistant as you would to a smart collaborator.

Useful phrases include:

• “I’m new to this. Define technical terms the first time you use them.”
• “Give me one concrete example before the theory.”
• “Pause after step three so I can try it.”
• “Show me what good looks like and what a common mistake looks like.”
• “Ask only questions whose answers would materially change the result.”
• “Tell me what you are uncertain about.”
• “Challenge my plan if the evidence points elsewhere.”
• “Keep a running decision log.”

You are allowed to interrupt, correct, disagree, or ask for a simpler explanation. That is not failing at AI. That is using AI well.

7. Set up your professional AI workspace

Your workspace should answer four questions immediately:

1. What am I trying to achieve?
2. What source material is trusted?
3. What is still in progress?
4. Which output is the current approved version?

Create this folder structure

In your Documents folder, create a folder named AI-Workspace:

AI-Workspace/
├── 00-Inbox/
├── 01-Learning/
├── 02-Projects/
├── 03-Templates/
├── 04-Reference/
└── 99-Archive/

• 00-Inbox: temporary downloads, screenshots, voice notes, and ideas to sort.
• 01-Learning: course notes, exercises, and your learning log.
• 02-Projects: one folder per real outcome.
• 03-Templates: reusable prompts, briefs, checklists, and formats.
• 04-Reference: trusted guides you expect to reuse.
• 99-Archive: completed or inactive work you do not want to delete.

Give every project the same shape

Create your first project folder at 02-Projects/My-First-AI-Project:

My-First-AI-Project/
├── README.md
├── brief.md
├── sources/
├── working/
├── output/
└── decisions.md

The files have simple jobs:

• README.md explains what the project is and how to use it.
• brief.md holds the goal, audience, constraints, and definition of done.
• sources/ holds approved inputs.
• working/ holds drafts and experiments.
• output/ holds deliverables ready to share.
• decisions.md records important choices and why they were made.

Use names that survive time

Avoid final-final-v2-really-final.docx.

Use:

2026-08-22-course-outline-v01.md
2026-08-22-course-outline-v02-reviewed.md
2026-08-23-course-outline-approved.md

Use dates in YYYY-MM-DD order. They sort correctly on every computer.

8. Create your AI Project hub

Inside your primary AI assistant, create a project or persistent workspace named:

AI Learning Lab

Add these project instructions:

You are my practical AI learning partner.

My current level: beginner.
My goal: become confident using AI for research, writing, operations, and building digital projects.

How to teach me:
- Use plain English and define new terms.
- Start with a concrete example, then explain the principle.
- Give me one action at a time for setup tasks.
- Never pretend an action succeeded; verify it.
- Separate facts, assumptions, and recommendations.
- When information may have changed, use current sources and cite them.
- Warn me before any action that publishes, pays, deletes, shares, or exposes private information.
- End lessons with a short practice task and completion check.

If your assistant does not offer Projects, keep the instructions in 03-Templates/AI-Learning-Partner.txt and paste them into a new chat when needed.

9. Privacy, safety, and verification

Before uploading anything, use the SAFE check:

Never paste these into an ordinary prompt

• passwords, one-time codes, API keys, or private keys;
• full payment-card or bank details;
• government identification numbers;
• confidential client or candidate information without explicit permission;
• private company documents into a personal AI account;
• data you would be uncomfortable seeing in the wrong workspace.

Match verification to risk

When researching, ask for direct sources. Open the important source. Check that it actually supports the claim. A citation is not decoration; it is a path you can audit.

10. Understand your live build environment

You are already using a live development workflow with this Dvans AI website. Here is what each piece does:

The live loop

Describe the outcome
        ↓
Inspect the current files
        ↓
Make one focused change
        ↓
Run checks
        ↓
Open localhost in Chrome
        ↓
Review what a real user sees
        ↓
Keep, revise, or undo

localhost is not the public internet. It means “this computer.” Your development server must keep running for the address to work. Closing the terminal or stopping the process stops the preview; it does not delete your files.

Four commands worth recognizing

npm run dev
npm run lint
npm run build
git status

• npm run dev starts the live local preview.
• npm run lint checks for suspicious code patterns and consistency problems.
• npm run build verifies that a production version can be created.
• git status shows files changed since the last saved Git snapshot.

You do not need to memorize commands. You need to understand their purpose and read the result before moving on.

11. Hands-on lab: direct your first live improvement

Your project is this course platform. Your first task is to improve one visible part of it without trying to redesign everything.

Step 1 — Choose one outcome

Pick one:

• make the homepage introduction feel more welcoming to beginners;
• add a “What you will learn” section;
• improve the mobile menu;
• add a progress indicator to this lesson; or
• rewrite one confusing paragraph in this module.

Step 2 — Write the brief

Create a prompt with BRIEF:

Background:
I am learning AI by improving a live course website. The current site is running locally.

Result:
Improve [one chosen area] so that [specific learner outcome].

Inputs:
Use the existing Dvans AI design and wording as the starting point.

Expectations:
Keep the purple/cyan dark visual language. Use plain English. Make one focused change. Preserve working behavior. Check mobile readability.

Format:
First explain the proposed change in three bullets. After I approve, update the relevant file and open the result in Chrome.

Step 3 — Review the plan

Before approving, ask:

• Is this still one focused change?
• What file will change?
• What could break?
• How will we know it worked?

Step 4 — Build and observe

Keep the development server running. Make the change. Refresh or allow hot reload. Look at the page as a learner, not as its creator.

Step 5 — Give evidence-based feedback

Avoid “I don’t like it.” Say what you observe and the effect:

The heading is clear, but the paragraph introduces four technical terms before explaining why they matter. Reduce it to one promise and move the terms below the call-to-action.

That feedback is actionable because it identifies the location, problem, and intended result.

12. Pro habits to start today

1. Define done before starting. If you cannot describe success, the AI cannot reliably optimize for it.
2. Keep source material separate from drafts. You should always know what came from evidence and what the AI generated.
3. Save decisions, not every conversation. Chats are working memory; your project files are durable memory.
4. Make focused changes. Small changes are easier to review, test, and undo.
5. Ask for uncertainty. A strong collaborator tells you where confidence is low.
6. Verify before publishing. Check facts, links, names, dates, permissions, and the actual visual result.
7. Use version history. Git is a time machine for project files; document history is the equivalent for writing.
8. Protect secrets. Public repositories and screenshots are common ways credentials leak.
9. Prefer repeatable workflows. A saved brief and checklist are more valuable than one lucky prompt.
10. Stay the editor. AI can propose; you decide what represents your name.

13. Knowledge check

Answer without looking back first.

1. What are the five parts of BRIEF?
2. When should you use a source-grounded notebook instead of an open-web research tool?
3. Why is localhost not the same as a published website?
4. Name three kinds of information you should not paste into an ordinary AI chat.
5. What is the purpose of the critique turn in the five-turn workflow?
6. What is the difference between a project folder and a repository?

Answer key

1. Background, Result, Inputs, Expectations, and Format.
2. When the answer should be grounded in a defined collection of trusted files or sources.
3. localhost is served from your own computer and works only while the local server is running; a published site is hosted for other people to access.
4. Examples include passwords/API keys, financial details, government IDs, confidential client data, or private records without permission.
5. To inspect the draft against a defined quality lens before revising it.
6. A project folder stores the files; a Git repository also tracks their change history and can connect to a remote host such as GitHub.

14. Your completion checklist

Mark this module complete when you have:

• chosen one primary AI assistant;
• created the AI-Workspace folder structure;
• saved the reusable BRIEF prompt template;
• created an AI Learning Lab project or saved its instructions;
• completed one five-turn prompt workflow;
• explained npm run dev and localhost in your own words;
• written a focused improvement brief for this live website; and
• recorded one lesson in 01-Learning/learning-log.md.

Use this learning-log entry:

Date:
What I built or learned:
What surprised me:
What I verified:
What I still do not understand:
My next small action:

Official references and optional practice

• ChatGPT Projects — OpenAI Academy
• Claude Artifacts — Anthropic Help Center
• Gemini Deep Research — Google Help
• NotebookLM sources — Google Help
• Getting started with Git — GitHub Docs
• GitHub Codespaces quickstart — GitHub Docs

Next: you will learn how AI systems receive data, generate outputs, use context, and fail—without unnecessary mathematics. That mental model will make every future prompt and tool choice more intelligent.