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

A VentureBeat study of 157 enterprises reveals a significant "evaluation gap," where organizations grant AI agents increasing autonomy despite having low confidence in the tests meant to ensure reliability. Half of the surveyed companies have deployed agents that passed internal evaluations but subsequently failed in production. While few trust current automated evaluation tools due to poor real-world alignment, two-thirds of organizations are either already deploying agents without human oversight or actively building pipelines to do so. To mitigate these risks, firms are increasingly investing in production observability and human-in-the-loop workflows, even as they move toward greater automation.

(Source：VentureBeat)