Coordination benchmarks predict AI tipping

Measuring AI by individual task benchmarks (e.g., score on a test) misses the economic value that comes when models are integrated into organizations and coordinate with humans and other systems. Small improvements in coordination or data plumbing can tip an industry from absorbing gains to reorganizing around AI — so policymakers and firms should track organizational‑level metrics, not only model capability tests. — Shifts where regulators, investors, and labor focus measurement and policy from model performance to organizational readiness and coordination, changing how we forecast AI’s economic effects.