Set Claude up right.
Use it like a professional from day one.

Most people come to Claude the same way they came to Google: they type a question and expect an answer. That's the wrong mental model, and it's why they get mediocre results.

Claude is a reasoning engine. It doesn't look things up. It doesn't browse the web during your session. It thinks — in context — using everything you give it in the conversation. If you don't give it context, it works with what it has, which is often not enough to produce work at a professional standard.

What Claude can do for operations professionals

Used correctly, Claude is a force multiplier for the kind of work operations professionals do every week:

• Build spreadsheets, trackers, and dashboards — with formulas, logic, and formatting
• Write and structure SOPs, procedures, and process documentation
• Draft executive briefs, status updates, and leadership memos
• Build revenue forecasts and scenario models
• Turn rough field notes into structured client-facing reports
• Translate documents into Spanish for shop-floor and field teams

What Claude cannot do

• Remember previous conversations. Each new conversation starts fresh, unless you use Projects (covered in Section 4).
• Browse the internet in real time. It works from training knowledge plus whatever you give it in the conversation.
• Verify facts from your operation. It doesn't know your numbers, your people, or your clients. You have to supply that context.
• Replace your judgment. It can do the work. You still have to review it.

Claude has a free tier. It's fine for experimentation. It is not sufficient for professional use, and I won't pretend otherwise.

Here's the practical difference:

Claude comes in three variants. Anthropic updates these regularly, but the principle is always the same: more capable models are slower and cost more usage. You pick based on what the task actually demands.

You switch models from the model selector at the top of any Claude conversation. The interface shows which model is active. Make a habit of checking before you start a high-stakes session.

The single most important feature for professional use is Projects. By default, Claude has no memory between conversations. Projects solve this.

A Project is a persistent workspace where you can:

• Upload documents that Claude can reference in every conversation within that project
• Set a system prompt — standing instructions about your role, standards, and preferences
• Keep related conversations organized in one place
• Build up a working knowledge base over time

How I set up projects for client work

For each client engagement, I create a Project and upload:

• A one-page "client context" document: the operation type, number of employees, main KPIs, any terminology specific to their industry
• Any existing SOPs or process documents I'll be building on
• My own firm's standards doc — formatting preferences, tone, document structure

Your first Project — set this up before the next section

1. Go to claude.ai and sign in
2. Click Projects in the left sidebar
3. Click New Project and name it something real — a client name, a department, or "My Operations Work"
4. Add a simple system prompt for now: "I'm an operations professional. Help me produce clean, professional deliverables — documents, trackers, and reports — to a high standard."
5. Upload one document you work with regularly — an existing SOP, a tracker, a process map. Anything.

Most people open Claude, type something, and see what happens. That's experimenting, not working. A productive session has structure. Here's mine.

• 1

  Define the output before you start

  Know what you're producing before you open Claude. "I need a production variance report for last week's Assembly A data" — not "I wonder if Claude can help me with reporting."

• 2

  Gather your context inputs

  What does Claude need to know to do this job well? Any data, standards, existing documents, or background that's specific to this task. Collect it before you start the conversation.

• 3

  Write a scoped prompt — not a question

  Set the role, give the context, define the task, and state the standard. We build this skill entirely in Module 02. For now: more context is almost always better than less.

• 4

  Review against a standard — not a feeling

  When Claude produces output, don't just read it and decide if it "feels right." Review it against a defined standard: Is it accurate? Does it meet the format? Would you hand this to a client without editing it?

• 5

  Iterate, don't restart

  If the output is close but not right, refine it in the same conversation. Tell Claude specifically what needs to change. "Change the variance column to show percentage, not raw numbers. Keep everything else." Don't start over unless the direction was fundamentally wrong.

What a scoped prompt looks like

Here's the difference between how most people use Claude and how this class will teach you to use it.

Write an SOP for our line changeover process.

ROLE: You are a process documentation specialist working for a manufacturing operations consultant.

CONTEXT: We run a mid-size packaging plant. Assembly Line 3 does a product changeover every Monday morning, from SKU 44A to SKU 62B. The process involves 4 operators and takes approximately 75 minutes when done correctly. It currently runs long because steps are not standardized.

TASK: Write a Standard Operating Procedure for the Line 3 changeover. Include: purpose, scope, required materials and tools, step-by-step procedure (with responsible party for each step), safety requirements, and a sign-off section.

STANDARD: Format it as a professional SOP document. Use numbered steps. Keep language direct — this will be printed and posted in the plant. Also produce a Spanish version immediately after the English version.

Same task. Completely different output. The second prompt gives Claude everything it needs to produce something close to usable in the first pass. The first prompt will get you a generic template you'll spend an hour revising.

You're going to run an actual session now, not a test. Pick a real task from your current work — something you'd normally spend 30–90 minutes on. A document you need to write. A tracker you need to build. A report you need to produce.

If nothing comes to mind immediately, use one of these:

• A one-page summary of a process in your operation that isn't documented anywhere
• A meeting agenda for your next team review or production meeting
• A draft of an email or memo you've been putting off

Run the session

1. Open your Project (or a new conversation if you're on free)
2. Write a scoped prompt using the structure from Section 5: Role → Context → Task → Standard
3. Review the output against a real standard — would you hand this to someone?
4. Make one round of refinements in the same conversation
5. Save the output — you'll reference it in Module 02