How do you automate weekly reports with AI?

A weekly report is an ideal automation target: same structure, similar inputs, and it recurs forever. The saving is not just the hour or two each week; it is the mental overhead of reassembling it.

For a full worked example, from the input folder to the review checklist, see the step-by-step guide Automate your weekly report with AI.

Step 1: Write down what you do today

For your last report, list:

• Inputs: which spreadsheets, exports, emails or dashboards you pull from.
• Steps: filtering, calculations, comparisons.
• Output: the format, sections and recipients.
• Judgment calls: what you comment on and why.

If you cannot describe it, you cannot automate it. The workflow-design pattern we teach is trigger, AI processing, human review, output.

Step 2: Standardize the inputs

Automation breaks when the inputs change shape. Save exports to a consistent folder with consistent file names and column headings. Put a copy of last week’s finished report in the folder as an example.

Step 3: Describe the job to an agent tool

Something like:

In this folder are this week’s sales export (sales.csv) and last week’s finished report (last_week.docx). Build a process that produces this week’s report in the same format. Calculate total revenue, change versus last week, top five customers, and any customer whose revenue dropped more than 20%. Write a three-sentence commentary. Show your calculations in a separate sheet so I can check them.

Claude Code, Codex and similar tools can build a script or workflow from this; see what Claude Code is and whether non-programmers can use it.

Step 4: Test against a known week

Run the automation on a past week whose correct report you already have. Compare each number. Fix mismatches by giving more precise rules. Do this for two or three past weeks before trusting it.

Step 5: Add the review step

Even a good automation should not send the report itself at first. Have it produce a draft for you to read. Look for: figures out of range, a commentary that ignores something you know happened, and formatting problems.

Step 6: Schedule or trigger it

Some tools can run on a schedule. Anthropic documentation says Claude Code scheduled tasks, formerly routines, run in the cloud on a schedule, on API calls, or on GitHub events. Anthropic says Claude Code desktop scheduled tasks suit recurring work such as morning briefings that pull from your calendar and inbox. Anthropic also says that by default, local scheduled tasks in Claude Code desktop run against whatever state your working directory is in, including uncommitted changes. Keep the report folder tidy. Microsoft says a Power Automate cloud flow can perform tasks such as sending a report in email. Microsoft says you can create a scheduled cloud flow using natural language in Copilot. Others you run manually with one command. Manual is fine at the start.

Step 7: Maintain it

Note changes in your data source, and update the instructions. Keep the instructions in a text file alongside the automation.

Things to guard against

• Silent data errors: a changed column name can shift figures without an error. Add checks such as row counts and totals.
• Confidential data flowing into an AI service: see whether it is safe to put company data in ChatGPT.
• Overtrust: keep a monthly manual spot-check.

If your data lives in Excel or Sheets, also see the Excel guide and AI with Google Sheets. In Learn AI Fast, everyone builds the same report generator from a written specification on provided data.

Sources

• Anthropic, Automate work with scheduled tasks (Claude Code): “Create scheduled tasks, formerly routines, that run in the cloud on a schedule, on API calls, or on GitHub events”
• Anthropic, Schedule recurring tasks in Claude Code Desktop: “morning briefings that pull from your calendar and inbox”
• Anthropic, Schedule recurring tasks in Claude Code Desktop: “By default, local scheduled tasks run against whatever state your working directory is in, including uncommitted changes”
• Microsoft Learn, Run a cloud flow on a schedule in Power Automate: “Create a cloud flow that performs one or more tasks such as sending a report in email”
• Microsoft Learn, Run a cloud flow on a schedule in Power Automate: “You can create a scheduled cloud flow using natural language in Copilot”