You can upload a messy spreadsheet to Claude, ask it to clean the data, and get back a dashboard with charts you can click through and share, all in one chat. No formulas, no coding, and it works on the free plan, though sharing a link needs a paid plan.
If you've ever been handed a client's "tracker" with merged cells, three date formats and a tab called FINAL_v2, this one's for you. Here's how to get from that sheet to a Claude dashboard, step by step, plus the checks I'd never skip.
To test it, I made a practice sheet that looks like a realtor client's listings research: 312 rows of made-up home listings with duplicates, mixed date styles, misspelled neighbourhoods, missing prices and a stray "average" row in the middle. No real client data went in. I ran everything in a normal Claude chat on my Max plan.
What Claude Can Do With a Spreadsheet
Two Claude features do the work here, and both are in Anthropic's help center.
The first is file creation and analysis. Claude runs code in a safe, closed-off space on Anthropic's side, so it can read the spreadsheet you upload, clean it, and give you back a new Excel file. You can upload CSV, TSV and other data files, up to 30 MB each. Anthropic lists it for every plan: Free, Pro, Max, Team and Enterprise.
The second is artifacts. An artifact is something Claude makes that opens in its own window beside the chat, like a document, a deck, or a dashboard. You can keep editing it and share it. Creating artifacts in a chat is also available on every plan.
One setting to check first. Artifacts and file creation both need the "Code execution and file creation" switch turned on. In newer versions of Settings, it's called "Create and edit files". On Free, Pro and Max, you'll find it under Settings, then Capabilities.
Claude Dashboard Examples: The Step-by-Step
1. Make a safe copy. Before anything touches AI, save a copy of the sheet and strip out what doesn't need to be there: customer names, emails, phone numbers. A dashboard about listings by month doesn't need anyone's personal details.
2. Ask Claude to describe the mess first. Upload the file and ask: "Before you change anything, list every problem you see in this data: duplicates, mixed date formats, blank rows, typos in category names." For my listings sheet, I swapped the last part for "typos in neighbourhood and property type names." This gives you a to-do list you can check, instead of fixes you can't see.
[ADD IMAGE: claude-problem-list.jpg]
Claude found every problem I'd hidden in the sheet: 6 duplicate listings, 5 blank rows, the stray average row, 18 missing prices and 5 prices with an extra zero. It even noticed the sheet mixed "Harbor View" and "Harbour View". Then it asked me three questions before touching anything, like whether "Single Family" should count as "House".
3. Clean it, and keep the receipt. Ask Claude to fix the problems and give you a cleaned Excel file, plus a short note on every change it made. You want the note, because your client will ask why a number moved.
In my test, the cleaned file came back with a change log sheet listing every fix, and the original file stayed untouched. Its average price, $489,704 across 277 listings, matched my own check to the dollar.
4. Ask for the dashboard with real questions. This is the step most people rush. Don't just say "make a chart." Say what the client wants to know. Here's what I asked: "Build an interactive dashboard from the cleaned data that shows new listings by month, the average list price by neighbourhood, the top five neighbourhoods by number of listings, and which neighbourhood's prices dropped since last quarter, with a filter by property type."
Anthropic's own tip for its dashboard feature says the same thing: say what you want to know, not just what to chart.
It took two follow-up messages. First, Claude showed me its plan and waited for my go-ahead before building anything. Then the "change since last quarter" column came out cut off on the right edge, so I asked it to fix that. The dashboard opened in its own window beside the chat, and the property type filter updated every chart on the page.
5. Check the numbers against the sheet. Pick three figures on the dashboard and find them in the original spreadsheet yourself. If the total for March on the dashboard doesn't match a simple sum in Excel, ask Claude to explain how it got there.
Claude told me straight out that it hadn't opened the dashboard to look at it, and asked me to check it over. Good thing I did. Harbor View showed a 29.6% price drop since last quarter, and Claude's summary said to treat that number with care. So I clicked the House filter, and the drop shrank to 10.4%. Harbor View simply had mostly houses listed from April to June, and mostly condos from July to September. If I'd sent the first number to a realtor, it would have looked like prices crashed.
I asked Claude to add one line under the table explaining the difference, and it did, with the numbers to back it up.
6. Share it. Artifacts start private. On Pro and Max, you can switch one to "Anyone with the link" or invite people by email. On Team and Enterprise, your client's admin decides whether links can go outside the company. Check who can see it before you send it, the same way you'd check a Google Drive link.
What About Claude's Dashboards Template?
You might see a "Dashboards" template in the Claude app. It's a different tool. It's in beta on paid plans (Pro, Max, Team and Enterprise), and it's built for data that lives in a company database or in a connected app like Salesforce. Claude writes the queries and each chart shows how it got its number.
For a spreadsheet someone emailed you, the regular chat steps above are what you need. If your client's data already lives in one of those systems, the template is worth a look.
What This Means for VAs
Reporting is one of the most useful things a VA can offer, and one of the most tedious to build by hand. With Claude, the cleaning and charting become a conversation.
But the value isn't the chart. It's that you asked the right questions, removed private data before uploading, and checked the totals before your client saw them. That's what makes a dashboard trustworthy, and it's still your job.
Start with a sheet you already know well, so you can spot a wrong number straight away.
If you build one, tell me what your client said about it.
