Messy spreadsheet cleanup is exactly the kind of grunt work AI handles well.

The transformation rules — normalize dates, unify currency, standardize deliverable text, enforce consistent casing — are structurally the same across all 1,200 rows. Once the logic is defined, it applies uniformly, making this highly automatable.

Most success criteria are crisp (ISO dates, consistent capitalization, unified currency), but the unlabeled USD/EUR mix is a genuine ambiguity that requires either a heuristic rule or a human decision before the agent can proceed confidently. That one gap prevents a 'high' rating.

The data lives in a single Google Sheet, which is easily accessible via the Sheets API or a direct export. No external data sources, live APIs, or special permissions are needed beyond read/write access to the file.

Errors are reversible if the original sheet is preserved or version history is enabled — a critical precaution. However, silent errors (e.g., wrong currency conversion, misclassified project status) could corrupt the Q1 business review if not caught, so a human spot-check is warranted.

The vast majority of this work is rule-based transformation with no taste, ethics, or relationship context required. The only genuine judgment call is resolving the currency labeling ambiguity, which can be handled upfront with a simple decision rule from the user.