Claude 101

Claude is an AI you talk to. You type, it types back. It can write, summarize, analyze documents, work with your files, and help you think through a decision.

It’s made by Anthropic. If you’ve used ChatGPT, Claude is the same kind of thing. Same species, different personality.

Three concepts that will save you a lot of frustration

One · It’s autocomplete, at scale. Claude predicts the next word, billions of times per response. It’s pattern-matching at a speed that feels like thinking, but it isn’t thinking the way you do. This is why Claude sounds confident even when it’s wrong. Remember this. It matters.

Two · It’s trained to agree with you. Claude is trained to be helpful. Side effect: if you say something false, it might nod along. If you ask “is my pricing page good?” it’ll find reasons to say yes. You have to push. Ask it to disagree. Ask what’s wrong. Ask for the strongest case against.

Three · It thinks in tokens, and it runs out of them. Claude reads and writes in chunks called tokens. Roughly, a word. Every conversation has a limit. Long chats eventually fill up the memory and Claude starts getting worse. When that happens, start a new chat.

Same species. You talk to both the same way. Different strengths.

Where Claude wins

• Writing voice. Less AI-flavored. ChatGPT sounds like a corporate memo. Claude sounds more like a person.
• Long documents. Claude handles 200-page PDFs without losing track.
• File work. The desktop app reads your local folders and works with your actual files.
• Multi-step jobs. Cowork mode runs tasks for minutes or hours while you do something else.

Where ChatGPT still wins

• Voice mode. Not even close.
• Image generation.Claude doesn’t make images. Use ChatGPT or Gemini.
• Deep research.ChatGPT’s deep research still edges out Claude’s.

Go to claude.ai. Sign up with your email. Three plans you need to know about:

• Free. Works in the browser. Limited messages per day. No Cowork. Fine for a week of testing.
• Pro · $20/mo. Best model, more usage, Cowork, Projects. Where most people land.
• Max · $100–$200/mo. Heavy usage. For people who hit the Pro ceiling every week.
• Team / Enterprise. For orgs. Admin controls, SSO, data not used for training.

My rule for picking

• Freeif you’re just deciding.
• Proif you’ll use it more than 3 times a week.
• Maxif Pro keeps running out or you’re running Cowork on long tasks daily. This is where I live.

Pay monthly, not annually. Test for 30 days. If you don’t reach for Claude by week three, cancel. You’ve lost $20, not $240.

The question every CMO, CEO, legal team, and IT lead will ask. Short version: it depends on the plan.

• Free and Pro plans.By default, Anthropic does not use your conversations to train their models. You can verify this in your settings. But: if you opt in to sharing conversations for research, that changes. Don’t opt in if you’re uploading anything sensitive.
• Team and Enterprise plans.Your data is never used for training. Period. This is the plan to pick if you’re uploading customer data, MMFs, pricing strategy, board materials, or anything under NDA.
• What Claude sees in a chat stays in that chat. It doesn’t remember you across conversations unless you’re using a Project or have memory turned on.

What this means in practice

• Uploading your MMF? Fine on any paid plan.
• Uploading customer contracts, PII, or regulated data? Team or Enterprise. Not Pro.
• Uploading competitor analysis, internal financials, or strategic plans? Team or Enterprise for peace of mind, Pro if you’re a solo founder and the risk is yours alone.

Send this link to your legal or IT team if they want the full terms: anthropic.com/legal/privacy. If they need SOC 2, HIPAA, or enterprise-grade controls, Enterprise is the plan.

This is where beginners get lost. “Claude” shows up in several places. Five of them matter. The rest you can ignore for now.

Browser Claude (claude.ai)

You type, you get an answer. Works on the free plan. Best for quick drafts, summaries, thinking out loud. If you’re brand new, start here.

Desktop Claude (Mac / Windows)

Same account as the browser, installed on your computer. The important difference: it can see your local files, and you unlock three modes — Chat, Cowork, and Code.

Cowork (paid plans only)

This is the real magic. Cowork is Claude running on your computer for minutes or hours, doing actual work while you do something else. You give it a task. It plans the steps, reads your files, produces outputs, and asks you questions along the way.

Concrete examples — especially powerful when pointed at your MMF:

• Point it at a folder of 80 customer interviews plus your MMF. Ask it to find quotes that prove your three core beliefs. Come back in 20 minutes with evidence for your next board deck.
• Feed it your last 10 sales call transcripts. Ask it to build a list of every objection and how the rep handled it — then rewrite the weak ones using your MMF’s language.
• Give it your homepage URL, your three top competitors, and your MMF. Ask where yours sounds the same and what three changes would sharpen your category distinction.
• Drop in a messy CSV export from your CRM. Ask it to clean it, flag duplicates, and give you a usable file.
• Hand it a SOW template and notes from a kickoff call. Ask for a first draft.

The difference between Chat and Cowork: Chat answers one question at a time. Cowork executes a project.

Claude Design (NEW · April 2026)

Launched six days before I wrote this guide. Claude Design turns text prompts into working prototypes, pitch decks, one-pagers, landing pages, and mockups. You describe what you want. It builds a first version you can click through. You refine it with comments or direct edits.

Why it matters for you:

• Your pitch deck, built from an outline, in minutes. Export to PPTX or Canva.
• A landing page mockup for a new offer, before you hire a designer.
• Product wireframes your PM can hand to your dev team.
• It reads your existing site or codebase and applies your actual design system, so the output looks like your brand, not AI slop.

It’s in research preview. Rough edges are real. But it’s already good enough to replace the “can you mock something up so I can see it?” step. Available on Pro, Max, Team, and Enterprise.

Claude Code (for your engineers)

Claude running inside a terminal. Built for developers. If you don’t code, you don’t need it.

But: if you have engineers, forward them this section. Claude Code is genuinely changing how software gets built in 2026. Your dev team should know about it. It connects directly to Claude Design, which means a prototype can get handed off to code with one command.

This is the section that separates people who dabble in Claude from people who get real leverage out of it.

A Project is a folder inside Claude that remembers context across every chat. You load documents into it once — your MMF, your brand voice guide, your customer list, sample content you like — and every chat inside that Project starts with all of it loaded.

Why this matters for your MMF

Without a Project, you’re pasting your MMF into every new chat. Exhausting. Inconsistent. Tokens burned on setup instead of work.

With a Project pre-loaded with your MMF:

• Every LinkedIn post draft already knows your category, your enemy, your three beliefs.
• Every sales email already knows how you talk about your ideal customer.
• Every board deck already knows your strategic narrative.
• Every brainstorm starts from your positioning, not a blank page.

Your Project becomes the persistent memory of your strategic story. You stop re-explaining. Claude stops drifting into generic SaaS voice. Your output stays 100% on-brand without you policing every sentence.

If we already built your AI Brand Twin

Open the Claude app. Click Projects in the sidebar. Find yours. Every chat you start inside it has your MMF, your voice, and your positioning already loaded. You’re not starting a chat — you’re talking to your AI Brand Twin. Just start writing.

If you have an MMF document but no Project yet

Do this, in order:

1. Open Claude (browser or desktop, either works).
2. Click Projects in the left sidebar. Hit New Project.
3. Name it your company name.
4. Upload your MMF document into the project knowledge.
5. In the custom instructions, paste something like: “You are a strategic content assistant for [Company]. Every output must reflect the voice, category, beliefs, and positioning in the uploaded MMF. Never drift into generic SaaS language. Before writing, reference the framework.”
6. Save. Open a new chat inside the project. Start working.

That’s it. Your AI Brand Twin is live. From here on, every chat you start inside that project is a conversation with a version of your company that knows your story cold.

Or tell us and we’ll build it for you.

Here’s a trick most people miss.

Your MMF controls what you say. The beliefs, the category, the enemy, the narrative spine. That’s locked.

Voice is separate. Voice controls how you say it. And once your AI Brand Twin is live, you can layer any voice you want on top of it without losing the story underneath.

Want your next LinkedIn post in Ryan Reynoldsvoice? Self-deprecating, meta, slightly unhinged? Tell Claude. It’ll write the same strategic argument but sound like a Mint Mobile ad.

Want a blog intro in Seth Godin voice? Short paragraphs. A provocative question. An aphorism. Same MMF beliefs underneath. Different delivery.

Want a founder letter in Anthony Bourdain voice? Sensory, no-BS, blue-collar respect. Same strategic narrative. Sounds nothing like the Seth Godin version.

This is why voice-anchoring is a superpower most founders never unlock. You spent weeks nailing what your company says. Voice is how you make it sound like a human wrote it, in a specific key, on a specific day.

How to actually do it

Two ways, both work:

Name a reference. Tell Claude: “Rewrite this in the voice of [person]. Keep every MMF belief and the strategic narrative intact. Change only the tone, rhythm, and word choice.” Works for famous writers, executives, comedians, anyone with enough public writing for Claude to pattern-match against.

Paste examples. Drop in three posts, essays, or emails by the voice you want to borrow. Tell Claude: “Write like this. Keep my MMF intact. Only borrow the tone and structure.” This works better than naming someone when the voice is lesser-known, or when you want to sound like yourself on your sharpest day.

You can also build up a library inside your Project. A folder called voices-i-like with samples labeled. Then your prompts get short: “Write the next LinkedIn post in the Bourdain voice.” Claude knows where to look.

One last thing

You can mix voices. “Write this in Bourdain voice but with Seth Godin’s paragraph length.” Works. Sometimes lands weird. Always interesting.

If you’re the CMO, marketing director, or the person who’ll actually operationalize Claude day-to-day, this section is for you.

The CEO reads this guide once and uses Claude for a few high-leverage tasks. You’re going to live in it. Here’s what that looks like.

Treat the Project as your content operating system

Every marketing team I’ve worked with that gets leverage out of Claude does the same thing: one well-loaded Project becomes the hub. Your MMF is the spine. Then you add supporting documents as you go:

• Your brand voice guide (or three great samples if you don’t have one)
• Your ideal customer profile (ICP) document
• A file of approved claims, metrics, and case study data
• Five to ten of your best-performing LinkedIn posts, emails, or blog intros (examples beat instructions, every time)
• A list of phrases you never want to see in your copy (looking at you, “leverage,” “synergize,” “empower”)

Every week, you add one more useful asset. In 90 days your Project has the institutional memory of your marketing function. In a year, it’s a moat.

Build a workflow, not a prompt library

Most marketing teams try to build a prompt library. It ages fast and nobody uses it. Instead, build workflows — the 5 to 10 repeatable motions your team runs constantly.

Examples:

• Weekly LinkedIn drop.Monday morning: give Claude the week’s theme, ask for 5 post drafts in your voice, pick 3, schedule them.
• Customer interview → content. Drop a transcript in. Ask for 3 quotes, 2 post ideas, and 1 blog angle, all aligned to the MMF.
• Campaign brief → draft assets.Give it the brief. Ask for landing page copy, email sequence, and 3 LinkedIn posts in one pass. Edit, don’t rewrite.
• Sales enablement. New objection shows up in sales calls? Ask Claude to draft a response card using MMF language. Ship to the team same-day.
• Competitive monitoring.Once a month, feed it three competitor homepages. Ask where they’re sharpening and where you’re drifting.

Once a workflow runs three times and works, it becomes a Skill. That’s when you stop doing the work yourself.

The role is changing. Lean in

Let’s be honest about what this does to the marketing function.

Claude doesn’t replace a good marketer. It replaces a mediocre one, and it makes a good one dangerous. The CMO who learns to orchestrate Claude across their team will outproduce a traditional department 3x. The CMO who treats it as a novelty will get outrun within a year.

Your job changes from “producing content” to “directing Claude to produce content that’s 90% right, then making the last 10% excellent.” That last 10% — the taste, the judgment, the strategic call on what’s worth publishing — that’s your actual job now. It always was.

If you’re at an agency

Everything above still applies, but harder. Your clients are going to figure out Claude within 18 months. The agencies that survive are the ones that already know how to use it better than the clients — bringing strategy, taste, and speed the client can’t match in-house. The agencies that try to hide it, or bill for work Claude could have done in an hour, won’t make it.

Five rules. That’s all you need.

One · Be specific. “Write me an email” gets you generic garbage. “Write a follow-up email to a client named Sarah who missed our Tuesday call. Friendly but firm. Four sentences.” Gets you something you can send.

The more specific your input, the more specific the output. This is the whole game.

Two · Give it examples.This is the single highest-leverage thing you can do. Paste three LinkedIn posts you wrote that you liked. Tell Claude: “write like this.” It will learn your voice faster from examples than from any instructions you could write.

(If your Project has your MMF loaded, this is already happening automatically. That’s the whole point.)

Three · Say what you want, not what you don’t. “Don’t be too formal” is weaker than “write it like a text to a colleague.” Tell it what to do, not what to avoid. The one exception: if Claude keeps doing something you hate, tell it you hate that thing.

Four · Start short. Add detail. Don’t write a 500-word prompt on your first try. Two sentences. See what comes back. Then: “make it shorter,” “change the tone,” “add a section on X.” Building up beats getting it right on the first shot. That’s why it’s called a chatbot.

Five · If it gets confused, start a new chat. Long conversations get messy. The context window fills up and Claude gets worse. When responses feel off, open a new chat and paste in the key context. A fresh start is free.

Non-exhaustive list. These are the ones that will move the needle for a B2B founder with an MMF in hand.

Content in your voice. LinkedIn posts, blog drafts, newsletter intros, sales emails, cold outreach. With your MMF loaded, every draft starts on-brand. You stop editing for tone and start editing for sharpness.

Brainstorming inside your positioning. New offer ideas, campaign angles, webinar topics, content themes — all generated from inside your category definition, not from a blank page. This is where most CEOs waste weeks and Claude saves them.

Writing internal docs. Board decks, investor updates, all-hands memos, customer communications. Your MMF ensures your strategic narrative carries through every document, not just the ones you personally touch.

Summarizing long documents. Drop a 50-page PDF in the chat. Ask for a one-page summary with page references so you can verify. Works on earnings reports, analyst research, contracts, customer interview transcripts.

Thinking out loud with you.Structuring a messy idea. Pressure-testing a decision. Claude is a good thinking partner if you push back on it rather than accepting the first answer. Ask it to argue against your plan. Ask what you’re missing. Ask it to steelman the other side.

Working with your actual files. On Desktop and in Cowork, Claude reads your local folders and creates files (Word, Excel, PowerPoint, PDF) directly on your machine. No copy-paste ping-pong.

Knowing the limits will save you from the most common disappointments.

Real-time information.Claude doesn’t know what happened today unless it’s searching the web. It will guess and sound confident doing it. For breaking news or current prices, use ChatGPT with search or Grok.

Precise math.Don’t use Claude as a calculator for anything that matters, unless it’s running code to compute the answer. It’s a language model, not a math engine.

Vague prompts.“Write something good” gets generic garbage. Claude needs specificity. You get what you give.

Being a source of truth.It sounds authoritative even when it’s wrong. Always verify facts, dates, quotes, and names before you send anything to a client or publish anything external.

Making images.Claude reads images. It doesn’t create them. Use ChatGPT or Gemini for that.

Every word you type, every file you upload, every response Claude gives gets chopped into tokens. Your usage limits are counted in tokens. Long chats hit ceilings because the context window is finite.

A page of text is roughly 500 tokens. A long chat can hit millions. Claude Code and Cowork sessions can burn through your monthly quota faster than you expect. Especially if you’re uploading large PDFs, working with codebases, or running Cowork on long tasks.

What this means in practice

• Start a new chat when the current one gets long. Fresh tokens, fresh Claude.
• Don’t upload a 300-page PDF if you only need 20 pages. Trim first.
• If you’re hitting limits on Pro every week, you’re a Max user pretending you’re not.
• Cowork and Design have separate usage quotas from chat. Use them without guilt.

Pick three. Do them this week. Each one teaches you something the last one didn’t. The examples below assume you have a Project set up with your MMF loaded.

1. Inside your Project, paste three LinkedIn posts you wrote. Ask Claude to write three new ones in your voice, anchored to your three MMF beliefs.
2. Inside your Project, ask: “Given our MMF, what are five campaign ideas for Q3 that would sharpen our category position?”
3. Upload your last 50-page investor deck. Ask for a one-page summary — then ask what it’s missing based on your MMF’s narrative.
4. Upload your last five meeting notes. Ask for a decisions log: what got decided, who owns what, what’s still open.
5. Inside your Project, paste a messy email draft you’ve been avoiding. Ask Claude to rewrite it in your voice, carrying the MMF’s framing.
6. Point Cowork at your Downloads folder. Ask what’s in there and what you can safely delete.
7. Connect Gmail. Ask Claude to summarize every email from your top three clients this month, flagging anything that needs a response.
8. Inside your Project, give it your homepage URL and ask: “Does our homepage reflect the MMF we built? What’s drifting? What’s missing?”
9. Feed it your last quarter’s P&L. Ask what patterns it sees that you might have missed.
10. Inside your Project, ask: “Write a 400-word investor update for the month covering [3 things]. Use our voice from the MMF.”

You don’t have to change anything. Don’t cancel ChatGPT. Just pick one path.

If we’ve already built your MMF and AI Brand Twin

1. Open the Claude desktop app.
2. Click into your Project.
3. Start a new chat.
4. Ask: “Based on our MMF, write a 300-word LinkedIn post I could publish tomorrow that sharpens our category position.”
5. See how close to publishable it comes back. Watch how little editing you need.

If you have your MMF but no Project yet

1. Sign up at claude.ai. Upgrade to Pro ($20).
2. Install the desktop app.
3. Follow the five-step setup in section 06 above.
4. Then run the same prompt: “Based on our MMF, write a 300-word LinkedIn post I could publish tomorrow.”

If you don’t have an MMF yet

That’s the bottleneck. Every tactic in this guide assumes you have one, because an AI Brand Twin built on a real MMF is the difference between “generic AI content” and “on-brand content on demand.” If you’d rather not wing it, we build Magnetic Messaging Frameworks in 90 days, ship your AI Brand Twin with them, and then run the Open Kitchen engine on top of it.

Greg Rosner · Founder, PitchKitchen · Strategic Messaging & Positioning for B2B Companies