Current benchmarks for occupational AI agents are scoped primarily by economic values, telling a replacement story. We introduce JobBench, which evaluates AI agents on the workflows that experts identify as high-priority for delegation, empowering humans based on their needs instead of replacing them with GDP value. JobBench covers 130 agentic tasks across 35 occupations. Each task is packaged as a workspace of heterogeneous reference files, requiring the agent to reason through the cluttered information streams of real professional work. Outputs are graded by a fact-anchored chain of rubrics, averaging 35.6 binary criteria per task. We evaluate 36 models; the strongest, Claude Opus~4.7 under Claude Code, reaches only 45.9 %. We hope JobBench shifts the community's target labour-market effect from replacement to enhancement: building agents that do what humans actually want delegated, not only what is most economically valuable.
@article{arxiv.2605.26329,
title = {JobBench: Aligning Agent Work With Human Will},
author = {Yuetai Li and Yichen Feng and Zhangchen Xu and Zixian Ma and Kaiyuan Zheng and Fengqing Jiang and Xinghua Sun and Rulin Shao and Zichen Chen and Yue Huang and Xinyang Han and Brian Lee and Kayla Xu and Shenglai Zeng and Hang Hua and Xiangliang Zhang and Basel Alomair and Ranjay Krishna and Luke Zettlemoyer and Pang Wei Koh and Bhaskar Ramasubramanian and Luyao Niu and Xiang Yue and Radha Poovendran},
journal= {arXiv preprint arXiv:2605.26329},
year = {2026}
}