Claude for Dummies

I have spent an embarrassing amount of my working life explaining Claude to people who were convinced, about ten minutes earlier, that AI was coming for their job, their house and possibly their dog. It rarely takes any of the three. What it does do, if you point the right part of it at the right problem, is quietly hand you back the better part of a day a week.

The trouble is that Claude has grown a lot of features, and most people meet two or three of them and assume that is the whole show. They open the chat box, ask it to write an email, nod politely, and close the tab. That is the equivalent of buying a house and living in the hallway.

So this is the tour. No jargon for the sake of sounding clever, no breathless talk about revolutions. Just what each part does, when you would genuinely reach for it, and the one mistake that stops most people getting value out of it. I have written it for the business owner reading this at 7am with a coffee, wondering if this is going to be another thing they pay for and never use. It does not have to be.

What is Claude, in one sentence?

Claude is an AI assistant built by Anthropic. You talk to it in normal English and it helps you think, write, analyse and, more and more, get real work done across your files and tools. It sits in the same category as ChatGPT and Google Gemini. The difference that matters for you is not which model is a fraction cleverer this month. It is how much of your actual work each one can reach and finish.

The quick map: which bit does what

If you read nothing else, read this table. Everything after it is detail.

Chat: the front door

Chat is the blank box you type into, and for most people it is the whole of their relationship with Claude. It is genuinely useful on its own: drafting, summarising, explaining a contract clause in words a human can understand, turning your half-formed rant into a clear email.

The mistake nearly everyone makes is treating it like Google. You type three words, get a decent-but-generic answer, and conclude the thing is overrated. Claude is not a search engine. It is closer to a very fast, slightly over-eager graduate who has read almost everything and just needs telling, clearly, what you actually want.

How to get the best from it

Give it the context you are carrying around in your head. Who is this for, what tone, how long, what does good look like. Paste in the real example. Ask it to have a go, then tell it what is wrong. The people who get gold out of Chat are not writing clever prompts. They are just being specific, and being willing to say "no, shorter, and lose the enthusiasm".

Projects: a workspace that remembers

A Project is a room you set aside for one job, one client or one recurring piece of work. It holds its own files, its own standing instructions and its own history, so every time you come back Claude already knows the context. You are not starting from a cold, blank page each Monday.

This is the fix for the single most annoying thing about AI, which is having to re-explain your world every single time. Put your brand guidelines, your last three reports and a note on how you like things done into a Project once, and every conversation inside it inherits all of that.

How to get the best from it

If you catch yourself pasting the same background into the chat more than twice, that is your signal. Make a Project. Drop the reference material in, write two or three lines of instructions about tone and format, and let it carry the context so you do not have to.

Artifacts and Design: things you can see and keep

When Claude builds something more substantial than a paragraph, it does not bury it in the chat. It opens it beside the conversation as an Artifact: a document, a spreadsheet, a slide deck, a dashboard, a working web page. You can see it, edit it, download it and share it, and you can ask for changes in plain English rather than fiddling with the file yourself.

Design is the visual cousin of the same idea. It is a proper canvas for the things you judge with your eyes: a mockup, a poster, a landing page, a one-pager, a social graphic. You describe what you want, it lays it out, and you nudge it around until it looks right.

How to get the best from it

Reach for an Artifact whenever the output is a thing rather than an answer. And do not accept the first version out of politeness. It is a draft on a canvas, not a stone tablet. "Make the header smaller, move the chart up, and use our navy" is a perfectly good instruction.

Skills: teach it your way, once

A Skill is a reusable instruction pack that teaches Claude to do one specific job the way you do it. Your report structure. Your compliance checklist. Your tone of voice. You write the rules down once, save them, and from then on every relevant piece of work follows them without you having to nag.

To prove the point: this article was written through a Skill that holds my tone of voice, which is why it sounds like me having a coffee and not like a pamphlet from a mid-tier consultancy. Once that Skill exists, I never have to explain "no em dashes, British spelling, sound like a human" again. Claude just does it.

How to get the best from it

Look for the task you do most often to a fixed set of rules. That is your first Skill. The dull, repeatable, "everyone here knows how we do these" job is exactly the one worth teaching it, because it is the one you never want to think about again.

Connectors: plug it into your actual tools

Connectors give Claude a live link to the software you already run: Gmail, Google Drive, Slack, your calendar, your CRM, and a growing list of others. Instead of copying and pasting your world into a chat window, Claude can read and act on the real thing where it lives.

This is the step where Claude stops being a clever writing partner and starts being genuinely useful inside your business. "Find the last three emails from this supplier and summarise where we landed" only works if it can reach your inbox. Connectors are how.

How to get the best from it

You grant each connection deliberately, and you can switch it off whenever you like. So connect one tool you use daily, give it a small real task, and see how it feels before you wire up the rest. You stay in control the entire time. Nothing is touched until you say so.

MCP: the plug socket behind it all

MCP stands for Model Context Protocol. If Connectors are the appliances, MCP is the shape of the plug and the wiring in the wall: the open standard that lets any tool connect to Claude in a consistent way. It is the reason the list of things Claude can talk to keeps getting longer without anyone having to reinvent the socket each time.

Here is the honest bit. As a business user, you will almost never touch MCP directly. It matters to you the way plumbing matters: quietly, and mostly when someone else has set it up well. You will never need to say the words "Model Context Protocol" out loud at a dinner party. But now you could, and people would leave you alone for the rest of the evening.

Cowork: where it stops answering and starts doing

This is the big one, and it is what the infographic at the top of this page lays out. Cowork is the mode where Claude stops handing you answers and starts doing the work. You give it a goal in plain English, it plans the steps, works across your files and connected tools, and hands back finished output for you to check. It runs in the cloud, which means it keeps going whether your laptop is open, shut or in a bag on a train.

The mental shift is small but it changes everything. Chat tells you how to do the job. Cowork does the job. Underneath it sits a handful of parts worth knowing by name.

Files and folders

You choose which folders on your computer Claude can see, and it reads, edits and creates real files right there inside them. No uploading, no downloading, no copying answers out of a chat window and pasting them into a document. It works where your work already lives.

Context and memory

Global and per-folder instructions set the ground rules, and a shared memory across Chat and Cowork means you explain yourself far less often. Tell it once that you write in British English, in your tone, and never want it to sound like a robot, and that sticks.

The built-in browser

Cowork has a browser inside it. Rather than handing you a numbered list of "now go and do this", Claude can open sites, fill in forms and finish web-based tasks itself, while you watch or get on with something else.

Scheduled tasks

You can set work to run on a schedule. It runs in the cloud and lands finished, whether or not you are at your desk, whether or not your laptop is even on. This is the part that starts to feel less like a tool and more like quietly having help.

Claude Code: for the technical few

One honest line for completeness. Claude Code is the version of Claude that lives in a developer's terminal and can read, write and change whole codebases. If you build software, it is superb and you already know why. If you do not write code, you can skip this one entirely and lose absolutely nothing. It is on the menu, not on your plate.

So which one do I actually use?

Here is the sensible order for a normal human running a normal business:

You do not need to understand all of it before you start. You need to give it one real task and see what happens. Most people reach a properly useful working level within a few weeks. That is the whole secret, and there is no clever trick hiding underneath it.

Frequently asked questions

The questions people actually type, and the answers, if you would rather ask an AI than read a table.

So pick one thing. The dull job you have been putting off, ideally. Give it to Claude properly, context and all, and see what Tuesday looks like on the other side. That is the entire trick. Everything above is just knowing which door to open.