Learn how to use Claude Desktop for quick data analysis without coding. A step-by-step setup guide for marketers.
Before Claude can work with your files, you need to give it a place to read and write. We'll walk you through how to create a single dedicated folder and connect it to Claude Desktop.
First, create a folder called Claude somewhere easy to find. On a Mac, open Finder and create it inside Documents. On Windows, do the same in your Documents folder or wherever makes sense. This is the one folder Claude will have access to. Think of it as a shared workspace between you and the AI.
Open Claude Desktop and head to Settings (click "Claude" in the Mac menu bar, or the settings icon on Windows). In the left sidebar, find the Capabilities tab and make sure the "Code execution and file creation" toggle is turned on. If you had to flip it, restart the app before moving on.
Still in Settings, scroll down to the Desktop app section in the left sidebar, then click Extensions, then Browse Extensions. Find and install the Filesystem Extension. This is what allows Claude to actually read and write files on your machine.
Once the extension is installed, you'll need to tell it which directory Claude can access. Add the "Claude" folder you just created. You'll also see a permissions option: choose "Always Allow" if you want a frictionless experience, or leave it on "Needs Approval" if you'd prefer to see exactly what Claude is doing before it touches any files. Hit Save, then restart Claude Desktop one more time.
That's it for setup. You now have a local AI assistant that can read your files, run code, and produce output, all on your machine.
Now for the fun part.
We'll take a real dataset - in this example a JSON export of hosted form data from a website - and turn it into a clean, grouped summary in Excel. No scripting. No pandas. Just conversation. This is what no-code data analysis actually looks like.
Download your data file and move it into the Claude folder you set up earlier.
Start a new conversation in Claude Desktop and type something like:
Start by first copying the file to Claude: 'www-sample-com_forms.json'
Claude will ask for file access permission the first time. Grant it, then let it work. You'll see it confirm that the file has been copied over and is ready to go.
Now start asking questions in plain English. A good opening move:
Look at the first few hundred lines of that file.
Claude will read the file, recognize the structure, and give you a summary of what's inside: field names, data types, how many records there are. This is your orientation step. You're getting a feel for the shape of the data without opening a single tool.
Once you know what you're working with, go after what you actually want to know:
Look at the forms in there, group them by name, and give me the names and counts of each form.
Claude will write and execute the code behind the scenes, then come back with a clean summary: a table of form names and how many submissions each one received. No SQL. No spreadsheet formulas. Just a question and an answer.
Happy with what you see? Ask Claude to make it permanent:
Make this a table in Excel.
Claude will generate a formatted .xlsx file and save it to your working folder. Open it up and you've got a polished spreadsheet ready to share, drop into a presentation, or build on further.
The workflow above took maybe five minutes from start to finish. But the real power is in what comes next. Once Claude has your data loaded, you can keep going:
For marketers, this means you can pull your own campaign insights without waiting on a data team or wrestling with a BI tool. Analyze form conversions, compare content performance across channels, or audit UTM data - all by just asking.
The only limits are your imagination and what Claude can express in Excel and TypeScript. So go nuts.
Most quick analysis tasks don't need a full BI tool or a data engineering pipeline. They need someone (or something) that can open a file, understand what's in it, and answer a few questions.
Where traditional BI tools require setup, configuration, and often a dedicated analyst, Claude Desktop gives you an on-demand data analyst that speaks plain English. It runs locally, works with your actual files, and doesn't require a license, training, or a dashboard nobody maintains.
That's not to say it replaces Tableau or Looker for enterprise reporting. But for the 80% of quick analysis tasks that marketers deal with - "how did that form perform last month?" or "which UTMs are actually driving conversions?" - Claude Desktop fills the gap perfectly.
Set it up once, and you've got an analyst on call whenever you need one.
Book a 1:1 demo and meet your new favorite (AI) employee.
