Create an Excel spreadsheet of one month of expense claims (March 2026) for a fictional 12-person marketing team, with about 80–120 rows and columns for claim ID, employee name, department, date, category (travel, meals, lodging, supplies, client entertainment), vendor, description, amount in USD, receipt attached (Y/N), and approver. Make it realistic: plausible vendors and amounts for each category, a few people who claim much more than others, weekends mostly empty, and a policy note on a second sheet (for example, meals capped at $75 per person per day, receipts required over $25, no self-approval). Hide exactly three mistakes that a careful reviewer could find from the data and the policy alone: one duplicate claim (same employee, vendor, amount and date under two different claim IDs), one policy breach (such as a $140 meal or an employee approving their own claim), and one data error (such as a date outside March or a total row that doesn’t match the sum). Keep the mistakes subtle: no highlighting, comments, or unusual formatting, and don’t place them all near each other. Put an answer key listing the three mistakes and their row numbers in a separate file, not in the spreadsheet.
Generated files, downloadable ZIP, and rerunnable pipelines all live here.
Build history, retries, and verification surfaces make the output easier to review with a team.
Start another run when you want to add features, fix issues, or generate a deployment package.
Did this project work out of the box?