The agent evaluation gap: Enterprise AI organizations have a reality-alignment problem, not a coverage problem — and most are shipping to production anyway

Across 157 enterprises, organizations are granting AI agents more autonomy while trusting the evaluations meant to gate that ...