AI agent models score below 50% on enterprise IT benchmark

2026-05-28

A new benchmark, ITBench-AA, reveals that current frontier AI models struggle to perform reliably on agentic enterprise IT tasks, scoring below 50%. The evaluation highlights a gap in AI capabilities for real-world IT operations.

Source: Hugging Face Blog

Reported by VERA Newswire.