PUBLIC_KEY_PEM before distributing.
Numbers in a spreadsheet mean less than numbers in a chart. This module teaches you to use LLMs to turn your business data into simple visual dashboards — as standalone HTML files you can open, share, and reuse without any external tools.
By the end of this module: you've taken a real dataset from your work, generated a standalone HTML visualization for it, and refined it with a second prompt to make the insight clearer. You've submitted both versions and noted what made the improved version more useful.
Most business data lives in spreadsheets. Sharing it means either handing over the file or taking a screenshot. Neither is ideal — one requires the recipient to have the right software, the other loses all interactivity and is hard to update.
A standalone HTML visualization is different: it's a self-contained file anyone can open in a browser. It shows the data as a chart or dashboard. It doesn't require Excel, Tableau, or any other tool. And you can generate one in a single LLM conversation by providing your data and describing what you want to see.
The LLM can't read your spreadsheet file — you need to paste the data into the prompt as text (numbers, labels, and structure). Keep datasets small for this exercise: a few rows and columns is enough. The technique scales, but the prompt needs to fit.
What kinds of business data work well for visualization? Almost anything with a comparison, trend, or distribution:
Monthly revenue by product, region, or rep. Week-over-week comparison. Pipeline stage breakdown.
Tasks by status (done, in progress, blocked). Hours by phase. Budget vs. actual by category.
Rating distributions. Yes/no breakdowns. NPS score breakdown. Satisfaction by category.
Response times. Error rates. Throughput. Headcount by department. Anything tracked over time.
Be specific about the chart type (bar, line, pie, table) and what insight you want to make visible. "Show me which products are growing fastest" produces a more useful chart than "show me the data." The LLM needs to know what the viewer should understand at a glance.
For this exercise, use a small sample of real data — or create representative numbers if your actual data is sensitive. The point is to practice the technique, not to process production datasets through an external tool.
Two parts. Part 1: describe your dataset, write a visualization prompt, generate the HTML chart. Part 2: improve it — add a second visualization, change the chart type, or make the insight clearer.
Pick something with enough structure to visualize — at least a few rows and two or three columns. Write a brief description of what the data represents.
Describe the data, what chart type you want, and what insight the viewer should get. Include the actual data in the prompt — paste the numbers as text, clearly labelled.
Save the generated code as a .html file and open it in your browser. Does the chart communicate what you intended?
What would make it more useful? Ask for a second chart, a cleaner layout, better labels, a summary number, or a different chart type for the same data.
Both HTML versions plus a brief description of the dataset and what changed between versions.
Describe the data, paste both versions of the generated HTML, and note what made the second version better.
Describe what the data represents, where it comes from, and what it shows. Include the actual data values as text (you don't need to paste the full spreadsheet — a representative sample is fine).
Paste the full prompt you sent to generate the visualization — including role, context, data, and instructions for the chart type and insight.
Paste the complete HTML code from the first version.
Did the chart communicate the intended insight? What was missing or could be clearer?
What did you ask the LLM to add or change? Paste the follow-up prompt.
Paste the complete HTML from the improved version.
What specific change improved the communication of insight? Would you use this in your actual work?
Your work is saved locally. Download to keep a copy for offline reference. Your work has been saved. Keep this file for offline reference.