Claude Code Tutorial: Your First Real Task From Start to Finish (No Coding)

Search “claude code tutorial” and most of what you find shows someone refactoring a Python module or wiring up a test suite. If you run a business and do not write software, that is close to useless. I kept meeting owners who had signed up, opened Claude Code, bounced straight off a tutorial written for engineers, and quietly decided the whole thing was not for them. So this is the tutorial I wish they had found. We are going to do one genuine task together, from a blank screen to a finished file, in plain English: taking a messy customer export and turning it into a clean, usable list. Nothing here asks you to write a line of code. You will just see the exact things to type and what comes back.

What you need before we start

Two things, and a spare five minutes. The first is a paid plan, because there is no free tier for the full tool. The entry plan is around 20 dollars a month and includes everything in this tutorial, and you do not need the pricier power user plans to begin. Anthropic bills in US dollars, so the pound figure floats a little with the exchange rate, and a card without foreign transaction fees keeps it tidy.

The second is a way to run it, and you have a choice here that matters more than people admit.

If the word terminal makes you tense, use the app and read no further about installation. If you are curious, the terminal route is a single paste, curl -fsSL https://claude.ai/install.sh | bash on a Mac, or irm https://claude.ai/install.ps1 | iex in Windows PowerShell, and then you type claude to begin. Either way you end up in the same place: a window waiting for plain instructions. For the rest of this tutorial I will assume you are in that window, however you got there.

The task we are going to do

Here is the job. You have exported your customer list from your CRM or your shop, and it is a mess. There are columns you do not need, the same customer appears two or three times, some email addresses are blank, and the formatting is all over the place. Doing this by hand is an afternoon of squinting and copying. We are going to hand it over instead. Save the file somewhere simple, make a copy of it so we never touch your only version, and call the copy customers.csv. That is the entire preparation.

Step 1: Ask what is actually in the file

Resist the urge to give orders straight away. Your first move is to let Claude Code look, so you both start from the same picture. Type something like this:

Look at customers.csv and tell me what columns it has, how many rows, and anything messy you notice in the data.

It will read the file and answer in plain English. You might hear that there are 812 rows, eight columns of which three are empty, around 40 duplicate email addresses, and a date column written three different ways. Already you know more about your own export than you did, and you have not done anything risky, because reading a file changes nothing.

Step 2: Give it the brief, then approve the plan

Now the real instruction. Notice how specific this is, because specific is what separates a good result from a vague one.

Make a clean copy called customers_clean.csv. Keep only these columns: Company, Contact name, Email, Last order date. Remove duplicate rows by email and keep the most recent order. Put any row with a missing or clearly invalid email into a Notes column flagged as “check email”. Leave my original file untouched.

Before it does a thing, Claude Code will show you the steps it intends to take and wait for you to approve them. This is the moment that makes the whole tool safe for a cautious owner. Read the plan. If it has understood, approve it and let it run. If it has misread you, say so in plain English and it adjusts. Nothing is written to a file until you have said yes.

Step 3: Check the result and ask for the proof

When it finishes it will tell you it is done. Do not just take its word. The habit that will serve you for years is asking for the evidence, so type:

How many rows did you start with and how many are in the clean file? How many duplicates did you remove, and how many rows did you flag for a bad email?

Now you can sanity check. If it started with 812 and the clean file has 760 with 40 duplicates removed and 12 flagged, the numbers add up and you can trust it. Open customers_clean.csv and glance down it. This whole step takes a minute and turns “the computer says it worked” into something you have actually verified.

Step 4: Save the recipe for next time

Here is the part that turns a one off result into a skill you keep. The brief you wrote in step two is reusable. Next month’s export will be just as messy, and you will not want to think it all through again. So save that instruction somewhere you can find it, a note, a document, wherever you keep useful things. The day a new export lands, you open the file, paste the same brief, and the job is a two minute paste rather than an afternoon. You have not just cleaned a list. You have built yourself a repeatable tidy up that costs you almost nothing to run again.

Adapt this Claude Code tutorial to your own work

The shape you just learned, look, brief, check, save, fits far more than customer lists. Once it clicks for one file it clicks for most of them. A few obvious next jobs, each one the same pattern with a different brief:

• Merge two exports. Hand it two spreadsheets and ask it to combine them on a shared column, keeping the rows that match and listing the ones that do not.
• Standardise a bookings or stock sheet. Ask it to put dates, prices, and names into one consistent format so the sheet stops fighting you.
• Pull a tidy list out of a long document. Point it at a PDF or a report and ask for the names, figures, or actions as a clean table.

If you are not sure which of your own files is the best place to start, my free AI Opportunity Scorecard gives you a quick read on where the repetitive hours are quietly leaking, which is usually a good clue to your best first task.

A few things that trip people up

Most first sessions stumble on the same small things, and knowing them in advance saves the frustration. Be specific in your brief, because “tidy this up” gives a worse result than the detailed instruction above. Keep one job to one conversation, since jumping between unrelated tasks in a single session confuses it. Always work on a copy while you are learning, so a misunderstanding costs you nothing. And if a long session starts repeating itself or contradicting earlier answers, its memory has filled up, so start a fresh session for the next job rather than letting one run all day.

None of these are technical. They are the same common sense you would use briefing a colleague, which is rather the point. The skill that makes Claude Code work for you is clear thinking about your own processes, and you already have that.

So if you have been put off by tutorials written for programmers, I hope this showed you the version that is actually for you. Do the one task, clean a real file you have been avoiding, and feel how quickly the fear turns into “what else can I hand over”. The reason I would rather teach you this than simply do it for you is that the skill, once it is yours, stays in your business and keeps paying you back, instead of being something you rent from someone else forever. When you want a hand picking and building your first proper automation on a real process, book an AI Opportunity Audit and we will do it together, or just get in touch and tell me the job that eats your week.