EnterpriseClawBench: Benchmarking Agents from Real Workplace Sessions
2026-06-22 • Computation and Language
Computation and LanguageSoftware Engineering
AI summaryⓘ
The authors created EnterpriseClawBench, a benchmark to test how well enterprise agents perform by using real work sessions from companies. They turned these sessions into 852 tasks with detailed setups and rules to evaluate the agents accurately. Because the data contains private company information, they only share how to build and test the benchmark, not the data itself. Their tests show that judging enterprise agents requires looking at many factors like cost, speed, and how well they deliver results, not just one simple score.
enterprise agentsbenchmarkingreproducible tasksprompt engineeringartifact deliveryevaluation protocoltool invocationskill transferCodexGPT-5.5
Authors
Jincheng Zhong, Weizhi Wang, Che Jiang, Kai Tian, Zhenzhao Yuan, Junlin Yang, Dianqiao Lei, Kaiyan Zhang
Abstract
Enterprise agents increasingly operate inside workspaces: they read heterogeneous files, invoke tools, and deliver business artifacts. We introduce EnterpriseClawBench, an enterprise agent benchmark constructed from proprietary, real-world agent sessions. Starting from a large archive of workplace sessions, the EnterpriseClawBench produces 852 reproducible tasks, each paired with recovered fixtures, rewritten prompts, role classes, skill subclasses, hard rules, and semantic rubrics. Because the sessions contain internal enterprise content, we do not release the benchmark data; instead, our reusable contribution is the construction and evaluation protocol. On EnterpriseClawBench, the best configuration reaches only 0.663 (Codex with GPT-5.5). These results show that enterprise agent evaluation must report harness--model combinations, artifact delivery, visual quality, cost, runtime, and skill-transfer behavior, rather than collapsing performance into a single score. Code: https://github.com/FrontisAI/EnterpriseClawBench