Dayjob: Tax – evaluating AI agents on 50 professional tax assignments

Surge AI released DAYJOB: Tax, a benchmark evaluating AI agents on 50 professional tax assignments requiring specialized expertise and judgment across eight tax practice areas. In testing, Claude Sonnet 5.5 achieved the highest pass rate at 10.8%, though all evaluated models struggled with complex tasks requiring professional judgment—such as identifying consequential deadlines that clients don't explicitly ask about. The benchmark uses 1,789 workspace files including tax returns, receipts, contracts, and other real-world documents to assess whether AI can perform at professional standards in tax preparation, compliance, and advisory work.