← All articles
EN NLFR
Tutorial Intermediate

Automate your monthly report with Claude

Stop spending hours on monthly reports. This tutorial shows you how to use Claude Cowork to turn raw data into a polished report — and how to automate the process.

AIClaudeCoworkproductivityautomation

Every month, the same ritual. Export data from three systems. Copy it into a spreadsheet. Make some charts. Write a summary. Format it in Word. Send it to management. Four hours of your life you’ll never get back.

What if Claude did 90% of that work?

This tutorial walks you through building a monthly report workflow with Claude Cowork — from raw data to polished document. And once it works, you can reuse it every month with minimal effort.

What we’re building

A monthly report that includes:

  • Key metrics with month-over-month comparison
  • Charts showing trends and breakdowns
  • Written analysis of what happened and why
  • A formatted Word or PDF document ready to send

Step 1: Gather your data

You probably pull data from multiple sources — a CRM, an accounting tool, a project management system, web analytics. For this tutorial, export whatever you have as CSV or Excel files.

A typical setup:

monthly-report/
├── CLAUDE.md
├── data/
│   ├── revenue-march-2026.csv
│   ├── projects-march-2026.csv
│   └── website-march-2026.csv
└── output/

Tip: Create a folder structure like this once. Every month, you drop the new data files into the data/ folder, keeping the same filenames.

Step 2: Write a CLAUDE.md for the report

This is where the magic happens. Your CLAUDE.md tells Claude exactly what the report should look like — every time, without repeating yourself.

# Monthly Report — CLAUDE.md

## Report structure
Generate a monthly management report with these sections:
1. Executive summary (max 5 sentences)
2. Revenue overview (total, by category, vs last month)
3. Project status (active projects, completed, at risk)
4. Website performance (visits, top pages, conversion)
5. Key takeaways and actions for next month

## Data sources
- revenue-*.csv: columns are Date, Category, Amount, Client
- projects-*.csv: columns are Project, Status, Budget, Spent, Deadline
- website-*.csv: columns are Date, Page, Sessions, Conversions

## Formatting
- Use professional language in Dutch
- Financial figures: € format with Belgian notation (€12.500,00)
- Percentages: one decimal (12,5%)
- Charts: clean style, company blue (#2A62FF) as primary colour
- Month-over-month comparisons: show absolute change AND percentage

## Output
- Create a single Word document (.docx) in the output/ folder
- Filename: rapport-[month]-[year].docx
- Also generate chart images as separate PNG files in output/

## Tone
Direct and factual. No filler phrases. Highlight problems clearly —
management prefers honest assessments over optimistic spin.

Step 3: Run the report

Open Cowork, select your monthly-report/ folder, and give a simple instruction:

Generate the monthly report for March 2026 using the data files in data/.

Claude will:

  1. Read your CLAUDE.md to understand the report structure
  2. Load and parse each CSV file
  3. Calculate the key metrics (totals, comparisons, changes)
  4. Generate charts using Python (matplotlib)
  5. Write the analysis sections
  6. Compile everything into a Word document
  7. Save it to your output/ folder

Depending on the data size, this takes 1-3 minutes.

Step 4: Review and refine

Open the Word document. The first time, you’ll probably want to adjust some things:

  • The charts don’t look right? Tell Claude: “Make the revenue chart a stacked bar instead of grouped. Move the legend to the bottom.”
  • The analysis misses something? “Add a paragraph about why consulting revenue dropped — it’s because Client X paused their project.”
  • Wrong formatting? “All table headers should be bold. Add a page break before each major section.”

Each correction takes seconds. And here’s the key: update your CLAUDE.md with these preferences. Next month, Claude gets it right the first time.

Making it repeatable

After the first month, the workflow is:

  1. Export data from your systems (5 minutes)
  2. Drop files into the data/ folder (1 minute)
  3. Open Cowork, select the folder, type “Generate the monthly report for April 2026” (10 seconds)
  4. Wait for Claude to generate the report (2 minutes)
  5. Quick review and any final tweaks (10 minutes)

Total: ~20 minutes instead of 4 hours.

Handling month-over-month comparisons

For comparisons to work, Claude needs access to the previous month’s data. Two approaches:

Approach A: Keep old data in the folder

data/
├── revenue-february-2026.csv
├── revenue-march-2026.csv
├── ...

Add to your CLAUDE.md: “For month-over-month comparisons, use the previous month’s file in the data/ folder.”

Approach B: Include a summary file After each month’s report, ask Claude: “Create a summary CSV with this month’s key metrics and add it to data/monthly-summaries.csv.” This gives you a running history.

Advanced: adding data from screenshots

Not all your data sources export to CSV. Some dashboards only show data on screen. Cowork can handle this:

Here's a screenshot of our CRM dashboard for March. Extract the key numbers
and include them in the report.

Claude reads the screenshot, extracts the numbers, and incorporates them into the report. Not as clean as CSV data, but surprisingly effective for systems that don’t offer data export.

Advanced: scheduled reports

If you’re on Claude Max, you can schedule Cowork tasks. This means you could potentially set up a report to generate automatically at the start of each month — provided your data files are updated.

See our intermediate guide for details on scheduled tasks.

Limitations

Data accuracy depends on your exports. Claude analyses what you give it. If your CSV has errors, the report will reflect those errors. Always verify the source data.

No live connections. Claude can’t pull data directly from your CRM or accounting tool (unless you set up connectors). You need to export the data first.

Complex calculations. For straightforward metrics (totals, averages, percentages, growth rates), Claude is excellent. For complex statistical analyses or custom financial models, you might need to provide the formulas in your CLAUDE.md.

Word formatting limits. The generated .docx files are clean and professional, but they won’t match a professionally designed template pixel-for-pixel. For important external reports, use Claude’s output as a draft and apply your company template in Word.

What’s next?