AutomationBench
AutomationBench is Zapier’s workflow automation benchmark, helping model teams, enterprise evaluators, and agent builders compare how well models complete multi-step SaaS automation tasks.
Tool overview
If your question is whether a model can actually complete enterprise automation work, AutomationBench looks relevant; if you want a tool to build automations directly, it is not that. More precisely, it is not Zapier itself and not a general LLM benchmark, but an AI eval/leaderboard centered on Zapier-style multi-step business workflows.
Based on the available evidence, its practical role is to compare models inside realistic workflow-like tasks. The sources repeatedly describe multi-step automation, common business functions, and setup details such as 657 tasks and 40 simulated SaaS environments. There are also references to 2-step workflows, explicit-instruction settings, small-model comparisons, and changing model scores over time. That supports its positioning as a benchmark for workflow automation ability, but does not prove that a leaderboard score maps cleanly to production outcomes.
On barriers and cost, the evidence mainly supports that it was publicly released; it does not reliably establish pricing, API fees, or a stable self-serve product surface.