How to Automate Your Work
with Claude · ChatGPT · Gemini

Strongest at long-form reports, complex analysis, and turning repetitive work into reusable Skills. Work through Beginner → Intermediate → Advanced to build up your toolkit.

1

Basic chat & prompt writing

Start a conversation on Claude.ai web or app. Structuring prompts as "Role – Context – Request – Output format" dramatically improves answer quality.

2

Choosing a model (Haiku / Sonnet / Opus)

Haiku for fast simple replies, Sonnet for everyday work like emails and social posts, Opus for high-stakes tasks like contract review or complex strategic analysis.

3

File upload & analysis

Upload PDFs, images, Word, Excel, or CSV files and ask for summaries, analysis, or translation.

4

Styles (conversation tone)

Save presets for formal, concise, or casual tone so you don't have to re-explain every time.

5

Searching past conversations

Search your chat history and save favorites for quick reference later.

Related videos

▶

YouTube

Claude Tutorial for Beginners — Step by Step (2026)

Everything a complete beginner needs to get started

▶

YouTube

FULL Claude Tutorial for Beginners in 2026 (25 min)

A complete beginner walkthrough in one sitting

1

Projects

Create a dedicated knowledge space for a specific task with reference material and custom instructions saved, so you don't have to re-upload the same files every time.

2

Artifacts

Generate working code, documents, web apps, and data visualizations in a side panel, and edit them right there.

3

Claude for Excel / PPT / Word

Create and edit spreadsheets with formulas, slides, and documents directly through conversation.

4

Connectors (MCP)

Connect Google Drive, Slack, Gmail, Linear, and more so Claude automatically pulls in the right data or sends messages when needed.

5

Claude Code basics

Even non-developers can write and run data-analysis scripts in natural language to handle large files or complex calculations.

Related videos

▶

YouTube

Claude AI Full Tutorial: From Basics to Agentic AI (2026)

Covers every major feature, free basics through agentic tools

▶

YouTube

Connect Claude Code to Google, Slack, and Reddit in 40 Min

Skills + MCP connectors, full walkthrough

▶

YouTube

FULL Claude Cowork Tutorial for Beginners (Zero to Pro)

Getting started with Cowork step by step

1

Building Skills

Save a repeatable task like "generate the monthly board report" — template, data sources, and emphasis points included — so anyone on the team can run it at the same quality.

2

Agent Teams

A lead Claude runs multiple Claude instances in different roles (market analyst, planner, developer, QA) in parallel. Each teammate works in its own context and collaborates via a shared task list and direct messages.

3

Cowork mode

An agentic workspace that handles file-system access, scheduled tasks, and folder organization across your whole desktop workflow.

4

Computer Use

Claude looks at your actual screen and controls the mouse and keyboard to automate other desktop apps (RPA-style tasks).

5

Custom automation via API/SDK

Embed Claude directly into internal systems to build fully custom automation pipelines.

Related videos

▶

YouTube

How to Properly Use Claude Code Agent Teams

Full live build of a multi-agent workflow

▶

YouTube

Claude Cowork Full Tutorial

Setup through real-world advanced workflows

▶

YouTube

FULL Claude Tutorial 2026 (Become a PRO)

Includes a full Claude automation course

1

Check your training-data setting first

Go to Settings → Privacy and turn off “Help improve Claude”. Since September 2025, consumer accounts are opted in by default. Claude for Work and API traffic are excluded from training by default.

2

Pseudonymize before you paste (GDPR core rule)

Under German data protection law (GDPR/BDSG), sending a customer's name, email, phone, or contract number to an external AI already counts as “processing personal data” and needs a legal basis. Replace real names with [Customer A] and contract numbers with [V-XXXX] before pasting; re-insert real values afterwards.

3

Separate personal and company accounts

Anything containing customer data belongs only in a company Team/Enterprise account covered by a DPA (data processing agreement). Keep your personal Pro account for learning and general research.

4

Audit files before Projects & connectors

Before uploading reference files to Projects, delete customer contact columns in spreadsheets and signatures/addresses in PDFs. Limit Google Drive/Slack connectors to the specific folders and channels you actually need.

5

Follow your company's rules

Classify data as public / internal / confidential / personal, and keep a whitelist of what may go into AI tools. If personal data slips in, delete the conversation and inform your data protection officer (DSB) immediately — GDPR fines can reach €20M or 4% of global revenue, and in Germany the company is liable.

Practical example — a pseudonymized prompt

Draft a polite reply to this customer complaint. (Personal data replaced) Customer: [Customer A] · Contract: [V-XXXX] · City: [City] Issue: double charge on the June invoice. Explain the refund process and timeline.

Related videos

▶

YouTube

AI Privacy Rules You Need to Know Now

What is safe — and unsafe — to share with AI tools

▶

YouTube

How To Keep Your Data Secure With AI Chatbots

Practical privacy settings walkthrough

1

Keep chats short — start fresh per task

Every reply reprocesses the entire conversation history, so long threads burn tokens exponentially. When the topic changes, open a new chat.

2

One well-scoped prompt beats five clarifications

A prompt with role, context, request, and output format costs far fewer tokens than a vague question followed by rounds of back-and-forth. Before sending, ask: “Could someone complete this task from this message alone?”

3

Attach only what's needed

Upload the relevant chapter, not the 100-page PDF; the one sheet, not the whole workbook. Put frequently reused material into Projects — cached project knowledge is cheaper than re-uploading every time.

4

Downshift the model

Summaries, translations, and email drafts are fine on Haiku/Sonnet. Save Opus for contract analysis and strategy work — usage limits drain faster on bigger models.

5

Store repeated instructions in Styles & Skills

Stop re-explaining tone and rules — save them once. To continue a long thread, ask for a 10-line summary of decisions so far and paste it into a fresh chat: the cheapest way to carry context forward.

Practical example — a token-efficient prompt

Role: financial analyst at a B2B services company Data: use only the 'DE' tab of the attached Q3 revenue sheet Task: top 5 items with the largest quarter-over-quarter variance + likely causes Format: one table + a 5-line summary (nothing else)

Related videos

▶

YouTube

Stop Wasting Tokens: The Art of Context Engineering

Why context — not longer prompts — saves your budget

▶

YouTube

The ADVANCED 2026 Guide to Prompt Engineering

Master the perfect prompt and cut wasted round-trips

Strongest at Excel modeling and browser automation (RPA-style tasks). Follow Beginner → Intermediate → Advanced.

1

Basic chat & model choice

Understand the speed/cost/quality trade-offs between GPT-5.5, 4o, and other models, and pick the right one for the task.

2

File upload & voice mode

Upload images or documents for analysis, or talk hands-free with voice mode.

3

Custom Instructions

Set your role, tone, and standing rules once so you don't repeat context every conversation.

4

Memory

ChatGPT remembers preferences and context from past chats and applies them automatically going forward.

Related videos

▶

YouTube

The Complete ChatGPT Tutorial for Beginners — 2026 Update

Daily-task basics through custom image generation

▶

YouTube

How to Use Custom GPT Chatbots (GPTs)

Create your own chatbot in minutes

1

GPTs (custom chatbots)

Build a chatbot dedicated to a specific task, upload knowledge files, and share it with your team.

2

Canvas

Co-edit a document in a side panel in real time, with version tracking and formatting tools.

3

Advanced Data Analysis

Upload data and have ChatGPT write and run code for cleaning, visualization, and regression analysis.

4

ChatGPT for Excel / Sheets

Officially launched May 2026. Build and update spreadsheet models in natural language, automating inventory management and budgeting.

5

Deep Research

Enter a research topic and get an automatically structured report built from dozens of web sources.

Related videos

▶

YouTube

ChatGPT for Data Analytics — Full Tutorial

From upload to insight

▶

YouTube

GPT Canvas Revolutionizes ChatGPT: Full Walkthrough

Hands-on with real-time co-editing

1

Agent Mode

In a virtual browser, ChatGPT navigates websites, logs in, fills forms, and downloads files like a human. Start it from the tools menu or type /agent. Tasks take 5–30 minutes (CAPTCHA-protected sites are a limitation).

2

Projects

Group related conversations and files into one workspace to manage long-running projects.

3

Custom GPT Actions

Connect external APIs to a GPT to build full automation that talks directly to internal systems.

4

Codex (dev automation)

A macOS desktop agent that automates writing and running code on your local machine.

Related videos

▶

YouTube

How to Use ChatGPT Agent Mode (Step-by-Step)

Browser automation, RPA-style tasks explained

▶

YouTube

The Ultimate ChatGPT Guide 2026: GPT-5.5 at Full Power

Advanced techniques beyond the basics

1

Turn off model training in Data Controls

Settings → Data Controls → disable “Improve the model for everyone”. For sensitive one-off questions, use Temporary Chat so nothing is kept in history.

2

Review what Memory stores

Memory can retain personal and company details across conversations. Check Settings → Personalization → Memory regularly and delete anything sensitive.

3

No customer data on Free/Plus accounts

For German companies, Free and Plus personal accounts are not suitable for processing customer data. Use a company Team/Enterprise account — it comes with a DPA, is excluded from training by default, and offers EU Data Residency (since 2025).

4

Make pseudonymization a habit

Replace real names with [Customer A], emails with [email], bank details with [IBAN] before asking — then re-insert the real values in the final document. This single habit removes most GDPR risk.

5

Be careful with shared GPTs & Actions

Never put customer data into the knowledge files of a GPT shared with your team. Get IT/DSB approval before connecting internal systems via Actions. If personal data slips in: delete the chat and report it immediately.

Practical example — rewrite before you paste

(Bad) "Write a reply to Thomas Müller (thomas.mueller@web.de, contract 4711) about his cancellation" (Good) "Draft a reply to customer [Customer A], contract [No.], about their cancellation request. Include the early-termination fee policy. Formal tone."

Related videos

▶

YouTube

Is ChatGPT Sharing Your Data? 6 Settings To Protect Your Privacy

Step-by-step privacy settings walkthrough

▶

YouTube

ChatGPT Privacy & Security Settings You Need to Change Right Now

Stop your conversations being used for training

1

New chat instead of endless threads

Long threads reprocess the whole history with every reply. Open a fresh chat per task; if you need continuity, carry over a short summary of the key decisions.

2

Custom Instructions kill repeated context

Save your role, tone, and output rules once in Custom Instructions — and stop spending tokens re-explaining your job in every conversation.

3

Upload only the relevant sheet or section

Give Advanced Data Analysis the one tab you need, not the whole workbook. Scope explicitly: “analyze columns A–F, 2026 data only” — less processing and fewer wrong answers.

4

Treat heavy features as a budget

Deep Research and Agent Mode consume a lot per run and have usage caps. Route summaries and drafts to lighter (mini) models, and fire the heavy features only once your question is fully formed.

5

Turn recurring work into GPTs

Put instructions and reference knowledge into a GPT so you never paste long prompts twice. On long projects, periodically ask for a “decision summary” to compress the context.

Practical example — a prompt that avoids round-trips

Role: copywriter on the marketing team Task: 3 newsletter blurbs based on the attached product sheet Constraints: max 60 characters each, include a CTA, formal tone Output: table (option / copy / CTA) — table only, no commentary

Related videos

▶

YouTube

Prompt Engineering Explained (2026) — Talk to AI Like a Pro

Turn basic responses into accurate results, fewer retries

▶

YouTube

Prompt Engineering for Beginners: Write Better AI Prompts

Get the answer you want on the first try

Strongest at real-time research and Google Workspace-based cross-department collaboration. Follow Beginner → Intermediate → Advanced.

1

Basic chat

Start a conversation in the Gemini app or web, and upload images or documents for analysis.

2

Gmail & Docs integration

On a Workspace Business plan, draft emails and summarize documents without installing anything or leaving the app.

3

Imagen-based image generation

Describe an image in text to generate it, and easily restyle it — "make it look like an animation."

Related videos

▶

YouTube

How to Use Google Gemini (Beginner Tutorial)

Writing, file analysis, and image creation

▶

YouTube

Google Gemini Tutorial For Beginners 2026 (Simplilearn)

Step-by-step foundations

1

Gems (custom agents)

Set a name, description, and clear instructions to create your own dedicated version of Gemini for a recurring task — similar to ChatGPT's GPTs.

2

Deep Research

Enter a topic and Gemini explores dozens of web sources and synthesizes a structured report asynchronously. Two versions: a fast preview model and a max-depth model.

3

Full Workspace integration

Auto-generate tables in Sheets, draft Slides decks, and have Meet auto-summarize meetings and email the notes.

4

NotebookLM integration

Store trusted source material in a notebook and Gemini treats it as a persistent knowledge base — no need to re-upload every time.

Related videos

▶

YouTube

Master Gemini in Google Workspace 2026

Complete tutorial for productivity

▶

YouTube

Master Gemini × Google Workspace (Free 1-Hour Course)

Gmail, Docs, and Sheets automation in practice

1

Gemini in Chrome — agentic browsing

A single prompt can complete multi-step web tasks like searching flights and booking a hotel. It reads up to 10 tabs at once, extracts and compares data, and you can take back control anytime with "take over."

2

Gemini Enterprise / Agentspace

Connect partner-built custom agents to automate end-to-end workflows.

3

Gemini API / Vertex AI integration

Build fully custom automation pipelines into internal systems.

4

Multi-tab research automation

An advanced research workflow that cross-checks data across multiple sites and auto-builds comparison tables.

Related videos

▶

YouTube

Gemini Gems + NotebookLM is INSANE

Advanced combo for research and documentation

▶

YouTube

Gemini in Google Workspace: Complete Tutorial

Plus NotebookLM & Gemini Enterprise

1

Turn off Gemini Apps Activity

By default conversations are stored and may be reviewed by humans to improve quality. Disable Gemini Apps Activity at myactivity.google.com and use Temporary Chat for sensitive questions.

2

Personal Google account ≠ company Workspace account

In Workspace Business/Enterprise, prompts are not used to train models and stay within your organization. Anything involving customer data belongs exclusively in the company account.

3

Know what the Gmail/Docs integration can reach

Gemini accesses mail and document content on request. If a document contains customer data, even “summarize this” is personal-data processing under GDPR — touch only what you need, only in the company account.

4

Pseudonymize before input

Swap real names for [Customer A] and addresses for [City] before asking. The same applies to source files you upload to NotebookLM — strip contact columns and signatures first.

5

Company guidelines & incident response

Agree on a per-team list of “data allowed into AI” and train new colleagues on it. If personal data was entered by mistake, delete it from the activity log and inform the data protection officer — under German law the company, not the employee, is liable.

Practical check — 3 seconds before you hit Enter

① Does this contain names, contacts, or contract numbers? → replace them ② Am I on my personal account or the company Workspace account? → customer data: company only ③ Is this data classified as "AI-allowed" by my company? → if unsure, ask the DSB

Related videos

▶

YouTube

Is Google’s Gemini Sharing Your Data? 6 Settings To Protect Your Privacy

Including how to stop training on your data

▶

YouTube

How to Manage Gemini Privacy Settings (Full Guide 2026)

Activity, review, and deletion settings explained

1

Mind the caps on heavy features

Deep Research (especially the max variant) and video generation have daily/monthly limits. Finalize scope, required sections, and output format before you launch — aim to run it once.

2

Flash vs Pro — pick deliberately

Email drafts and quick summaries run fine on Flash models, which drain limits far more slowly. Reserve Pro-class models for complex analysis and long documents.

3

Save recurring instructions as Gems

Build Gems like “weekly report writer” or “German business email polisher” — repeated instructions stop costing tokens and time.

4

Use NotebookLM for repeated questions on the same material

Instead of re-uploading files into every chat, register them once as NotebookLM sources. It answers strictly from your sources — efficient for repeated Q&A.

5

Short chats, specific requests

One prompt stating role, data scope, and output format beats several vague ones. Start a new conversation when the topic changes to avoid reprocessing history.

Practical example — one-shot Deep Research

Topic: AI adoption in German mid-sized manufacturing (2024–2026) Must include: ① adoption statistics ② top 5 use cases ③ 3 case studies ④ EU AI Act impact Output: report with table of contents, tables preferred, cite sources

Related videos

▶

YouTube

Prompt Engineering = Talking to AI the Right Way

Be specific, give context, guide the output

▶

YouTube

Learn Better Prompt Engineering in 7 Minutes

The essentials for efficient day-to-day prompting