A walkthrough of the Claude Academy course Claude 101 (Start your journey track, 13 lessons + quiz). Each lesson is summarised with my own tables and examples; take the course for the full lessons. All the courses are listed in the Claude Academy course index.
Note: Detailed content expanded from course outline and public lesson pages; verify against lessons. Module grouping is inferred from the course recap; plan availability and menu names change frequently.
Prerequisites: A Claude account on any plan (Free, Pro, Max, Team, Enterprise). No prior AI experience needed. Some features need paid plans: Cowork and Research (Pro, Max, Team, Enterprise); Enterprise Search (Team, Enterprise). On Team/Enterprise, admins control designs, decks, and docs.
Audience: Anyone starting with Claude at work — new users wanting a guided path to confident daily use, and existing users wanting a structured tour of projects, artifacts, skills, connectors, Enterprise Search, and Research.
Module 1 — Getting started (Lessons 1–4)
What is Claude?
- An AI assistant / thinking partner, trained with Constitutional AI principles to be safe, beneficial, transparent (helpful, harmless, honest).
- Capable of summarization, search, writing, Q&A, coding, reasoning; steerable on personality, tone, behavior.
- Available on web, desktop, mobile for all plans; conversations, projects, memory, preferences sync across devices.
- Strengths: writing, research/analysis (200K+ token context ≈ 500+ pages; up to 1M on paid plans with supported models), coding, problem-solving with Thinking, learning (Learning mode guides your reasoning).
| Surface |
What it is |
| Claude.ai (web/desktop/mobile) |
Main chat experience — focus of this course |
| Claude Code |
Agentic coding tool; edits files, runs commands, commits |
| Claude Tag |
Claude in Slack |
| Claude Design |
Turns ideas/sketches into interactive prototypes |
| Claude for Microsoft 365 |
Sidebar in Excel, PowerPoint, Word, Outlook |
Your first conversation with Claude
- Talk to Claude like a coworker: natural, concise, conversational. You bring context and expertise.
- Prompt framework (from the 4D AI Fluency “Description” competency):
| Element |
Questions |
| Set the stage |
Your role, objectives, relevant context |
| Define the task |
What action — write, analyze, build? |
| Specify rules |
Style, tone, format, examples |
I’m the marketing lead at an indie streaming startup, preparing a Series A pitch deck. Research the independent film streaming market — trends, competitor positioning, growth opportunities. Use current web research with citations; structure as a professional report up to 5 pages with an executive summary, market analysis, competitive landscape, and growth opportunities.
- Add context: upload PDF, DOCX, CSV, TXT, PNG/JPEG (text + visuals analyzed); enable tools (web search, connectors like Google Drive).
- Models: Opus for most complex tasks; Sonnet as recommended everyday default. Changing model starts a new chat. Extended thinking helps complex analysis but adds latency.
- Iterate: follow-up questions, feedback, redirect, or restart; pencil icon edits and resubmits a prompt.
- Personalize: Instructions for Claude (Settings > Account) apply to every chat; Memory saves role/preferences/decisions (on by default for Free/Pro/Max; off by default on Team/Enterprise, Owner must allow); Skills encode reusable workflows.
Getting better results
| Challenge |
Fix |
| Too generic |
Add audience, role, constraints |
| Wrong length |
State it (“two paragraphs”, “under 100 words”) |
| Wrong format |
Show an example / describe structure |
| Confident but wrong |
Verify key facts; ask for sources/confidence; enable web search |
| Wrong tone |
Describe tone plainly; give a sample |
- Iteration mindset: first drafts are starting points; specific feedback (“cut the first two paragraphs…”) beats vague; know when to start fresh.
- 4D Framework for AI Fluency (Dakan & Feller):
| D |
Meaning |
| Delegation |
Deciding what work goes to humans vs AI |
| Description |
Communicating clearly with AI |
| Discernment |
Critically evaluating outputs and process |
| Diligence |
Using AI responsibly, transparently, accountably |
- Simple evals: gather 5–10 examples of a recurring task → write test prompts → compare outputs (key info? tone? gaps?) → refine prompts or identify where human review is essential.
- Delegation–diligence loop (Rio example): test AI on past data you already analyzed; refine; delegate only what it reproduces reliably. Validation builds confidence but doesn’t remove accountability.
How you’ll work with Claude on your desktop
| Shape of work |
Where it lives |
Use when |
| Turn by turn (thinking partner) |
Chat |
Answers shape your next question; you want to stay in it; quick tasks |
| Handing work off |
Cowork |
Multi-step; output is finished files in your folders; spans tools; scheduled/background |
| Building software |
Code tab |
Read/write/test code in a repo (Local or Cloud) |
- Chat desktop extras: quick entry (double-tap Option on Mac), screenshots/window sharing, dictation, desktop connectors.
- Cowork extras: local folder access (saves back to folder), scheduled tasks, subagents, projects, Claude in Chrome browser use, computer use (research preview, Pro/Max), plugins (Customize → Plugins). Asks before consequential actions.
- Code tab: permission settings like manually approve / accept edits / plan.
Module 2 — Organizing your work (Lessons 5–7)
Introduction to projects
- Self-contained workspaces with their own chat histories, knowledge base, instructions, and memory.
- Create: Projects → + New Project → name, description (for humans; Claude doesn’t see it), visibility.
- Instructions apply to every chat (context, process, tone, requirements; can automate workflows like “When I upload a transcript, summarize with this template”).
- Knowledge: PDF, DOCX, CSV, TXT, HTML, Google Drive. Name files descriptively. Chat uploads stay out of project knowledge. Context isn’t shared across chats unless added to knowledge.
- RAG mode auto-enables near the context limit → searches knowledge, up to ~10× capacity.
- Team/Enterprise sharing: Can view, Can edit, Owner; share with people or whole org; “Shared with me” tab.
- Best practices: start focused, keep knowledge current, write clear instructions, reference documents by name.
Creating with artifacts
- Artifacts = standalone outputs (doc, deck, design, dashboard, prototype) in a side window; saved in the Artifacts tab for later editing/sharing. Say “Create this as an artifact” to force one.
| Tool |
Creates |
Exports |
| Claude Design |
Landing pages, one-pagers, UI mockups (editable canvas, design systems) |
.pptx, .pdf, .html |
| Claude Slides |
Presentations |
.pptx, Google Slides, .pdf |
| Claude Docs |
Living, co-editable rich-text docs; charts from connectors (snapshots) |
Google Docs, .docx |
- Also: charts, flowcharts, interactive pages, code — all plans including Free.
- Artifacts (open/update in Claude, shared by link) ≠ file creation (downloadable Word/Excel/PPT/PDF).
- Edit three ways: edit directly, comment for Claude, just ask.
- Sharing: private by default; view/comment/edit; Pro/Max can share public link; Team/Enterprise links stay in-org unless Owner enables External sharing.
- Tips: ask for the deliverable, be specific, describe the end user, build at the end of a conversation, refresh data on request.
Working with skills
- Skills = folders of instructions, scripts, resources loaded dynamically for specialized tasks.
- Anthropic Skills (Excel, Word, PowerPoint, PDF creation — auto-invoked) vs Custom Skills (your workflows).
- Enable: Settings > Capabilities → Code execution and file creation on (org-controlled on Team/Enterprise) → Customize > Skills → toggle.
- File execution: upload .xlsx/.pptx/.docx/.pdf; Claude creates updated versions (not in-place); may need “Allow limited network access.”
- Security: only install trusted skills; review external ones; custom skills private until shared.
- Create via conversation: “I want to create a skill for…” → answer questions → upload references → Save skill (or Customize > Skills > Add > Upload skill).
|
Projects |
Skills |
| Purpose |
Store knowledge (the what) |
Define processes (the how) |
| Best for |
Long-term context, team collaboration |
Repeatable, multi-step workflows |
| Persistence |
All chats in the project |
Applied when invoked |
Module 3 — Expanding Claude’s reach (Lessons 8–10)
- Connectors let Claude search, read, and (where allowed) act in your tools. Powered by MCP (“USB-C for AI”).
- Two types: web connectors (Gmail, Notion, Slack, Asana, Linear, Stripe…) and desktop extensions (local files, browser control, native apps; Desktop app Settings > Extensions).
- Find: connector directory, Customize > Connectors, or + > Connectors. One entry can cover multiple apps (e.g., Atlassian Rovo → Jira + Confluence). Custom connectors possible.
- Setup: Connect → authenticate → grant permissions → test (“Can you access my [tool]?”). Team/Enterprise: Owner adds the connector first.
- Security: scoped/toggleable permissions; Claude sees only what you see; revocable anytime; trusted sources only.
Enterprise Search (Team/Enterprise)
- “Ask {Org Name}” sidebar project — a pre-built org-wide project for finding/synthesizing internal knowledge, with Anthropic-configured instructions; always cites sources.
- Uses: catching up, policies/processes, research, onboarding, project tracking.
- Admin setup: Ask Your Org → Set up → required Documents (Drive/SharePoint) + Chat (Slack/Teams) connectors, email optional → name → Finish. Users then authenticate individually.
- Safety: only shows what you can already access; conversations private; data not separately indexed/stored.
Research for deep dives (paid plans)
- Agentic multi-step investigation: plan → many searches building on each other → synthesize → cite. Takes minutes; can span hundreds of sources; works with Thinking and connected integrations.
| Need |
Use |
| Comprehensive, multi-source, cited report |
Research |
| Quick fact from 1–2 sources |
Web search |
| Deep reasoning, no external info |
Thinking |
| Internal company knowledge |
Enterprise Search |
- Enable: + → Research (requires web search).
- Prompt tips: specific goals, desired structure/sections, constraints (budget, timeline, geography), ask Claude to help draft the prompt; steer integrations (“pull context from my Google Drive”).
Module 4 — Putting it all together (Lessons 11–13)
Claude in action: use cases by role
- General: status reports, feedback pattern analysis, brand guidelines skill.
- Sales: battle cards, deal prep, sales reports · Marketing: campaign analysis, cross-platform content · Finance: models, investment memos, inherited spreadsheets · HR: onboarding guides · Legal: discovery timelines · Research: literature review plans, verifying statistics.
Other ways to work with Claude
| Tool |
Best for |
Where |
| Claude.ai |
General tasks, research, writing, files |
Web, desktop, mobile |
| Claude Code |
Software dev, codebases, git |
Terminal, IDE, browser |
| Claude Cowork |
Multi-step tasks, docs, file organization |
Desktop (+ web/mobile per plan) |
| Claude Tag |
Team collaboration in context |
Slack |
| Claude Design |
UI prototypes, design exploration |
Any conversation + claude.ai/design |
| Claude for M365 |
In-place editing across Office files |
Excel, PowerPoint, Word, Outlook (beta) sidebars |
| Claude in Chrome |
Web research, inbox, browser automation |
Chrome sidebar (paid plans; asks before high-risk actions) |
What’s next
- Start simple: pick one recurring task this week, try it, iterate. Claude is a collaborator, not a replacement.
- Further: AI Fluency, AI Capabilities and Limitations, Claude Code in Action, Introduction to Claude Cowork, Use Case Gallery.
Quick-Review Flashcards
- Q: Three elements of a good prompt? → A: Set the stage, define the task, specify rules.
- Q: The 4Ds of AI Fluency? → A: Delegation, Description, Discernment, Diligence.
- Q: Where do you set preferences for every conversation? → A: Settings > Account > Instructions for Claude.
- Q: Chat vs Cowork vs Code? → A: Turn-by-turn thinking / handing off work / building software.
- Q: What happens when project knowledge nears the context limit? → A: RAG mode — Claude searches files, expanding capacity up to ~10×.
- Q: Three project permission levels? → A: Can view, Can edit, Owner.
- Q: Artifact vs file creation? → A: Artifacts live/update in Claude and share by link; file creation produces downloadable files.
- Q: Projects vs Skills? → A: Projects store knowledge (what); skills define processes (how).
- Q: What must be enabled for Skills? → A: Code execution and file creation.
- Q: What protocol powers connectors? → A: Model Context Protocol (MCP).
- Q: Research vs web search? → A: Research = multi-step, multi-source cited report over minutes; web search = quick fact.
- Q: Can Enterprise Search expose data you can’t normally see? → A: No — it respects your existing permissions in each tool.
Practice Ideas
- Rewrite a vague request you made recently using the stage/task/rules framework and compare outputs.
- Run a 5–10 example eval on a recurring task (e.g., weekly email) and note where Claude needs more context.
- Set Instructions for Claude and review what Memory has saved about you.
- Classify last week’s Claude requests as Chat, Cowork, or Code work.
- Create a project for an ongoing workstream with instructions and descriptively named knowledge files.
- Turn a chat analysis into a doc artifact, then “turn this doc into a deck”; practice commenting for Claude.
- Build a custom skill for your team’s report format via conversation.
- Connect one tool (calendar, Drive, Slack) and ask a question that requires it.
- Run a Research report on a competitive question with a specified structure; spot-check citations.
- Try one other surface (Claude in Chrome, Excel add-in, Claude Tag) on a real task.