How to Use Claude AI: The Operator's First Week, From Account to Daily Workflow

Most operators use Claude for a week the way they would use a new hire for a week: badly, then not at all.

The pattern is familiar. An account gets opened, three questions get asked, one answer is generic, and the tab closes. Six months later the same operator is paying a contractor to do the work the tool would have done in an afternoon, had the first week been run like an onboarding instead of a test.

This guide is that onboarding. It follows Anthropic's own documentation, as of October 2026, from what Claude is to the five features that turn a chat window into a working system: documents, Projects, artifacts, Skills and connectors. One feature a day, each installed on a real piece of your business.

By Friday the tool knows your offer, holds your documents, has built something you can put in front of a client, and runs one procedure you used to do by hand.

What Claude is, in Anthropic's words

The Claude Help Center describes it plainly: "Claude is a large language model (LLM) built by Anthropic. It's trained to be a helpful, honest, and harmless assistant with a conversational tone." It runs on the web at claude.ai, as a desktop app for Mac and Windows, and on iOS and Android.

On how to talk to it: "The best approach is to speak to Claude like you would a coworker or friend - naturally and conversationally." Hold that against the operator's instinct to type keywords into it like a search engine. It is a colleague who reads documents, drafts, edits, and reasons, and it responds to a brief, not a query.

One mechanical detail matters from day one. The Help Center notes that "The model you're currently chatting with is displayed either below your text input (on web and desktop) or at the top of your screen (on mobile)," and that the same menu "controls three settings: which Claude model you're chatting with, how much effort it puts into each response, and whether it uses thinking," per the article on model, effort and thinking settings. You will use that menu more than any other control.

Choosing a plan

As of October 2026 the pricing page lists Free, Pro, Max, Team and Enterprise. Free is described as "Free for everyone," Pro "For everyday work," Max "For people who work with Claude all day," and Team "For teams of 2 to 150." The comparison table shows Projects on Free as "Up to 5."

The Help Center adds the constraint that shapes a free trial week: "While using the free Claude plan, there is a session-based usage limit that will reset every five hours." Run the first week on Free if you want, but expect to hit the ceiling on the day you upload documents. The full breakdown of plans and which one an operator needs is in the Claude pricing guide; this article does not repeat it.

The first week, one feature a day

Each day installs one capability on one real asset. Do not practise on hypotheticals. Use the client, the offer and the procedure you actually have.

Day one: the first three tasks

Start where the Help Center's own troubleshooting article starts. Its guidance on unhelpful answers says: "Claude works best when you give it clear and specific instructions," and asks you to "Pretend you are giving these instructions to someone with no background knowledge about what you are asking." It also tells you to "Break down complex requests into substeps."

1. Rewrite one email you sent this week. Paste it and ask for a version half the length that keeps every commitment. Compare. You will learn how it edits.
2. Summarise one long document you have been avoiding. A contract, a report, a competitor's offer page. Ask for the five points that change a decision you have to make.
3. Think through one decision out loud. Describe the choice, the constraints and what you are afraid of. Ask it to list the assumptions you are making. Turn effort up for this one.

Write each request as a brief: the task, the context, the format you want back. The prompt engineering guide covers the eight techniques Anthropic documents; the short version is that vague asks get generic answers.

Day two: your own documents

Claude becomes useful to an operator when it reads your material rather than the internet's. The file upload article lists the supported documents, including PDF, DOCX, CSV, TXT, HTML and XLSX, and the limits: "File size: 500MB per file," "Up to 20 files per chat," and "PDFs are limited to 1000 pages." It adds that "Claude analyzes both text and visual elements (like images, charts, and graphics) in PDFs of 100 pages or fewer," and that for non-PDF documents "Claude extracts text only from these files."

Upload three things: your current offer document, your last ten client notes exported as text, and one spreadsheet of results. Then ask questions you could not answer quickly yourself. Which clients mention the same obstacle. Where the offer promises something the notes say rarely happens. What the spreadsheet shows month over month.

The article notes that "Files can be uploaded to individual chats or uploaded to a project's Files section for persistent reference across conversations." That second option is tomorrow.

Day three: a Project for one client or one offer

Projects are where the week turns from testing into a system. The Help Center defines them: "Projects allow you to create self-contained workspaces with their own chat histories and knowledge bases." They are "available to all users, including those with free Claude accounts," and "Free users can create a maximum of five projects," per the article What are projects?

Build one for your main offer. Upload the offer document, the brand voice notes and two examples of your best client-facing writing. Then set the instructions. The setup article explains that project knowledge is used "across all of your chats within that project," and warns: "Context is not shared across chats within a project unless the information is added into the project knowledge base." Put the standing context in the knowledge base, not in a chat.

The instructions field is where the role lives: who Claude is for this project, what standard it holds, what format it defaults to. Write it once. The Projects guide has five setups for coaches and operators.

Day four: an artifact you can put in front of someone

The Help Center's definition of an artifact is the operator's definition of a deliverable: "An artifact is anything Claude makes for you that you'd put in front of someone: a design, a deck, a document, a dashboard, or a small interactive tool." It appears beside the conversation, and you can edit it and share it. One setup note from the same article: "Artifacts require Cloud code execution and file creation to be turned on in Settings > Capabilities (Free, Pro, Max)."

Build one inside the Project from day three. A pricing calculator for your offer, with the client's inputs and your margin shown. Or a one-page onboarding checklist a new client can tick through. Ask for it in plain language and iterate in the artifact, not in a new chat. The artifacts guide lists seven worth building.

Day five: a Skill for your most repeated procedure

By Friday you will have typed the same instructions more than once. Anthropic's Skills documentation names the problem: "Unlike prompts (conversation-level instructions for one-off tasks), Skills load on demand, so you don't have to repeat the same guidance across conversations," per the Agent Skills overview.

The Help Center article on creating custom skills states that "Skills are available for users on Free, Pro, Max, Team, and Enterprise plans," that the feature "requires code execution to be enabled," and describes what the best ones do: they "Solve a specific, repeatable task" and "Are focused on one workflow rather than trying to do everything."

Pick the procedure you ran most this week. Writing the post-call summary. Turning notes into a client email. Drafting the weekly review. Write the steps as you would for an assistant, with one example of a finished output, and save it as a Skill. The Skills guide walks through the file format and five Skills every operator should have.

The sixth feature, when you are ready

Connectors come after the week, not inside it. The Help Center explains that "Connectors let Claude access your apps and services, retrieve your data, and take actions within connected services," and the permission model that makes them safe to try: "Claude inherits each person's permissions from the connected service," per the connectors article. Calendar and documents first. The protocol behind them is explained in the Claude MCP guide.

The habits that make it compound

The week installs features. These habits decide whether they compound or decay.

• One chat per task. A chat that drifts across five jobs produces worse work on all five. Open a new one inside the right Project.
• State the output format every time. A table, an email, a numbered checklist, prose paragraphs. The format you name is the format you get.
• Ask for the reasoning when the stakes are real. For a pricing decision or a hire, ask it to show its assumptions before its answer, and raise the effort. The Help Center notes that "Low and Medium work well for routine tasks and stretch your usage further," so save the high setting for decisions.
• Batch the small things. The usage guidance says: "If you have multiple related tasks or questions, group them in a single message," and that "Content in projects is cached and counts less against your limits when reused." Projects are cheaper as well as better.
• Keep the standing context in the Project, not in your head. Every time you explain your offer again in a chat, that explanation belongs in the knowledge base.

This is the same rule that governs training. A programme you have to re-plan every morning does not get followed. A programme that is written down, with the loads decided, gets executed. The Project and the Skill are the written programme. The daily chat is the session.

Mistakes beginners make

• Testing it on trivia. Asking what it knows about your industry tells you nothing. Give it your documents and ask what they show.
• Typing keywords. It is not a search box. Write the sentence you would say to a capable colleague.
• Accepting the first draft. The second message, the one that says what was wrong with the first, is where the quality comes from. The Help Center lists "Give Claude feedback" as a core principle for a reason.
• Never changing the effort setting. Routine drafting at high effort burns usage; a pricing decision at low effort burns money.
• Letting it decide what you want. It will produce something plausible for any vague request. The operator's job is to know what good looks like before asking. The Stoic review that closes each day is the practice that keeps that judgement sharp.

Frequently asked questions

How do you start using Claude AI?

Open an account at claude.ai, on the desktop app, or on the mobile apps, and write to it the way the Help Center suggests: "like you would a coworker or friend - naturally and conversationally." Start with three real tasks on day one: rewrite an email you sent, summarise a document you have been avoiding, and think through one live decision with the effort setting turned up.

Can you use Claude for free?

Yes. As of October 2026 the pricing page lists a Free plan described as "Free for everyone," alongside Pro, Max, Team and Enterprise. The Help Center notes that on the free plan "there is a session-based usage limit that will reset every five hours," and the comparison table shows Projects on Free limited to five. Paid plans are covered in the Claude pricing guide on this site.

What files can you upload to Claude?

The Help Center lists PDF, DOCX, CSV, TXT, HTML, ODT, RTF, EPUB, JSON and XLSX documents, plus JPEG, PNG, GIF and WebP images. Chat uploads allow 500MB per file and up to 20 files per chat, PDFs are limited to 1000 pages, and visual elements are analysed only in PDFs of 100 pages or fewer. Non-PDF documents are text-extracted only.

What is the difference between a Project, an artifact and a Skill?

A Project is a workspace with its own knowledge base and instructions, so standing context is uploaded once and used across chats. An artifact is a deliverable Claude builds beside the chat, such as a document, dashboard or small tool you can edit and share. A Skill is a saved, repeatable procedure that loads when the task appears, so you stop retyping the same instructions.

How long does it take to get good at using Claude?

One working week, if each day installs one feature on a real asset: documents on day two, a Project for your main offer on day three, an artifact on day four, a Skill for your most repeated procedure on day five. The habits that then compound are one chat per task, a stated output format, higher effort only for real decisions, and standing context kept in the Project.

Run the onboarding, not the test

A tool that is tested gets abandoned. A tool that is onboarded onto real work becomes part of how the business runs. Five days is enough to tell the difference, provided every day touches a real client, a real offer or a real procedure.

What you have at the end is leverage, in the operator's sense: the same hours producing more decided, finished work. The leverage only holds if the operator holding it knows what good looks like, which is why the discipline comes before the tool.

The free 5-Day Stoic Operator Challenge installs that discipline in the same five days: a training anchor each morning, a Stoic review each evening, and a day built from decided blocks. Run the two weeks side by side, and by Friday both the operator and the tool will know what they are for.