Turn Business Data Into Dashboards With AI

Most businesses already have the data they need sitting in Excel sheets scattered across random folders. In this walkthrough I show you how to turn that data into interactive HTML dashboards using Claude Cowork, without manually building a single chart. We generate mock receipts and sales records first, so you can follow along with zero privacy concerns before pointing this at your real numbers.

What you need

• The Claude desktop application, with the Cowork mode selected
• A folder with business data on your computer: Excel files, reports, even images. For this walkthrough we generate mock data instead
• If you work in a regulated environment, a plan for data retention (more on that below)

1. Open Claude Cowork and point it at local files (0:08)

We are using Claude Cowork for this build. Cowork is available through the Claude desktop application: download it, then click the button that selects Cowork. What makes it fantastic for this use case is that it works with local files on your computer. If you have a lot of Excel sheets just laying around in some random folder, Cowork can read them directly, no uploads or exports needed.

2. Generate realistic mock business data (0:31)

Since I cannot show you real client data, the first step is asking Claude to fabricate some. The Faker library is a Python library that replicates very realistic data, which is exactly what we need for a demo. Here is what I sent:

Hey, Claude, can you please generate fake data about receipts and also just sales information using the Faker library in Python for me? Thanks. I want about one hundred entries for each. It doesn't matter what the structure is.

This takes a couple of seconds to run. When it comes back, you have one hundred mock receipts and one hundred mock sales records, structured and ready to analyze. Remember this is only mock data for the demo. The exact same workflow applies to your actual business files.

3. Sort out privacy and data retention first (1:16)

Before you run this on real business data, think about privacy and data retention. If you are working in a regulated environment, contact the Anthropic sales team. They can set you up with zero days data retention, so nothing you process sticks around. Do this before the dashboards, not after.

4. Let Claude plan the dashboards with you (2:08)

Instead of describing every chart myself, I flip the conversation around and make Claude interview me:

Hey, Claude, can you now ask me some questions regarding data visualization so we can build some dashboards together to keep track of those receipts and also those sales processes?

This is the move that makes the whole workflow work. If you have a folder full of reports, Excel files, or images, Claude Cowork and Claude Code are now very good at analyzing the data and doing all the structuring and filtering for you. You just answer questions about what you want to see.

5. Choose KPIs, trends, and filters (2:42)

Claude asked whether I wanted an interactive HTML page, and that is what we build today. I asked for two separate dashboards and went for everything on the menu: KPIs, time series trends, and filters, so I could show you the full result.

One thing to know: generating the dashboards takes a few minutes, because an interactive HTML dashboard is actually code, and code generation takes time. How long depends on how big your data is. The good news is you only do this once. After the first build, you can set it up so the dashboard updates every time you add more data.

6. Walk through the finished dashboards (3:41)

In my run, Claude Cowork took about two minutes to generate both dashboards. The receipts dashboard gives a breakdown of everything I asked for: how much we are spending per store, spend rate by state, a breakdown by payment method, and daily spend over time. In two minutes we analyzed an entire Excel sheet, and this scales to sheets with one thousand to even ten thousand rows.

The sales dashboard is the same story: completely usable output from a single prompt. This is one of the best use cases I know, because it gives you a 360 view of what is happening inside your company at any point in time. You know where money is being spent and where money is coming in. If you work in marketing, the same approach gives you marketing analytics dashboards and real insights on top.

Pitfalls and tips

• Dashboard generation is not instant. Expect a couple of minutes whenever Claude is writing code, and longer for bigger datasets
• Build once, reuse forever: after the first generation, have the dashboard refresh when new data lands instead of regenerating from scratch
• Regulated industry? Talk to the Anthropic sales team about zero days data retention before touching real data
• Mock data first is a great habit. Validate the dashboard structure on fake records, then swap in the real files

Where to go next

The natural next step from here is turning these static HTML dashboards into live applications with Live Artifacts, which I cover in the following video. If you want to keep building business systems with AI, start with Build a CRM with Claude Code, then see how to feed your pipeline with content in Turn Calls into Content with AI.