BENCHMARKS
Tests anyone can run, check and cite.
Each benchmark publishes its protocol, fixtures, mock services and scoring rules at a versioned address before any result is accepted. Every published result shows all of its runs, failures included, with the evidence behind them.
VERSION 1.2 · PROTOCOL PUBLISHED · NO RESULTS YET
Enterprise Autonomy Benchmark for Governed AI Workflows
Can an AI application coordinate work across systems, follow a human approval decision, and leave enough evidence to reconstruct what happened?