Repeatability
High
Schema mapping, date normalization, and fuzzy deduplication are structurally identical operations regardless of the specific data. This is a classic ETL pattern that agents execute reliably.
Ambiguity Tolerance
Medium
Column remapping and date standardization have crisp success criteria, but 'same client' identification requires a confidence threshold decision — some near-matches will be genuinely ambiguous and need human sign-off.
Data & Tool Availability
High
The agent needs only the three spreadsheet files, which are presumably shareable. No live APIs, credentials, or external systems are required to complete the core task.
Error Cost
Medium
Incorrectly merging two distinct clients or splitting one client into duplicates corrupts the master record, but the source files remain intact and the output is reviewable before any downstream system is updated — making errors recoverable.
Human Judgment Required
Medium
Most of the work is mechanical, but edge cases — same company name with different addresses, name changes, subsidiaries — require a human to decide the business intent behind the merge rule.