Benchmarks / Enterprise workflows

Enterprise workflows

Multi-step work across enterprise software, email, and chat.

Part of Office and administrative support.

• TheAgentCompanyby Carnegie Mellon UniversityTask level · Tasks resolved

  42.9%TTE-MatrixAgent + DeepSeek-V3.2 · 10 Nov 2025

  Emerging
• WorkArenaby ServiceNowTask level · WorkArena-L1 success rate

  90.3%IpaziaHPA-Gemini-3-flash-preview · 7 Mar 2026

  Saturated