DAYJOB: Tax — Benchmarking AI Agents on Professional Tax Work
Summary
DAYJOB: Tax benchmarks AI agents on professional tax work using 50 environments and 1,789 files to test end-to-end performance, professional judgment, and handling incomplete information. The article presents results, describes the scope of tax practice covered, and emphasizes that current frontier models still struggle to identify consequential issues and take full ownership of client work.