Module 01 · Lesson 1

What Claude Is, Where It Fits, and How It Thinks

Claude is a reasoning assistant, not a search engine

Claude is an AI assistant built by Anthropic. The fastest way to understand it: Claude is good at the kind of thinking a smart, well-read colleague would do if you sat them down with your problem, your documents, and a coffee. Drafting, summarizing, comparing options, sketching plans, reviewing your work, walking through trade-offs, writing code, explaining things, all the parts of knowledge work that involve reading, thinking, and producing language.

It's not a search engine pulling current facts off the internet (though it can search when you ask). It's not a database with perfect recall. It's not a spreadsheet that does math for you (though it can write code that does). And it's not a person, it has no career, no consequences, and no skin in the game beyond doing well in this conversation.

A useful analogy: Claude is a brilliant intern with no memory

Imagine an exceptional intern who has read every book and document you've ever shown them, but only for the duration of the current conversation. They can think, reason, write beautifully, code, and synthesize. But the moment you close the chat, they walk out the door. Next time, they show up fresh. So the question becomes: how much context do you want to hand them, and what exactly do you want them to do?

The four jobs Claude is genuinely good at

After teaching hundreds of people, the same pattern keeps repeating. Claude shines at exactly four kinds of work:

1. Reasoning: comparing options, weighing trade-offs, walking through "if-then" logic, planning steps.
2. Writing: drafting, rewriting, summarizing, translating tone, generating variants.
3. Decision support: turning messy inputs into structured outputs (memos, tables, decision trees, frameworks).
4. Delegation: taking a goal and producing a sequence of work, including code, documents, and plans.

Notice what's missing: pulling live data, executing payments, or replacing human judgment in regulated decisions. Those are not what Claude is for, at least not without proper supervision and connectors, which we'll cover in later modules.

How Claude "thinks", the three pieces that matter

Whenever you send a message, Claude is working with three things at once:

• Context window: everything in the current conversation: your messages, Claude's replies, attached files, project instructions. Claude can read all of it, but only what's in the window. Anything outside doesn't exist for Claude.
• Instructions: the role, task, format, and constraints you give it. Vague instructions = vague outputs. There's no way around this.
• Reasoning: Claude can "think" step by step before answering, especially on complex tasks. Asking for reasoning explicitly often produces dramatically better results.

That's it. There's no fourth secret thing. Master those three and you've mastered 80% of working with Claude.

How to turn any business question into a Claude task

This is the workflow you'll use over and over. It's deliberately simple, that's the point.

1. Name the actual goal Not "I want help with marketing." Try "I want a one-page positioning brief for a B2B SaaS launching in February." Specificity is the multiplier here.
2. State the role and audience Tell Claude who it's playing ("Act as a senior brand strategist") and who the output is for ("for a non-technical CEO"). The same goal with different audiences becomes wildly different work.
3. Hand over context Paste relevant background, current copy, competitor names, target customer notes. If it lives outside the chat, Claude doesn't know it. If you skip this, Claude makes generic guesses.
4. Define the output format Bullet list? Markdown table? 3 versions, short/medium/long? Decision memo with a recommendation at the top? This is the single biggest fix when outputs feel "off."
5. Add constraints Word count, tone, things to avoid, must-include points. Constraints help, not hurt, they remove the infinite possibilities Claude would otherwise consider.
6. Ask for reasoning when it matters "Walk me through your thinking before you write the final version." For complex decisions, this often makes the answer noticeably better, and easier to trust.
7. Review, refine, finalize Treat the first reply as a draft, not an oracle. Tell Claude what to change. The second turn is usually the one that ships.

Three prompts that work on day one

Copy any of these. Replace the variables in orange with your specifics.

5-minute walkthrough you can recreate

Open Claude side by side with this lesson and follow along. Time stamps are rough.

Before vs. After: a real example

A founder asked Claude for "a pitch deck outline." Here's what changed when she added structure:

Five workflows from this lesson's teaching cards

Each card below is a real task pattern from the course's 1,750-card library. Use them as seeds for your own work.