Enterprise AI Agents Outpace Verification, Survey Finds
Companies are deploying increasingly autonomous AI agents, but confidence in automated testing is collapsing. Half of enterprises have experienced customer-facing failures from AI that passed internal evaluations.
A new VB Pulse survey reveals a significant gap in enterprise AI evaluation: increasingly autonomous AI agents are causing more production failures, while trust in automated testing processes is collapsing.
According to the June 2026 survey of 157 enterprise respondents, half of companies have deployed an AI agent or LLM feature that passed internal tests but still caused a customer-facing failure. One in four enterprises experienced this issue more than once.
The trend is not slowing development: 66% of respondents already permit some production deployments without human review or are building systems for it. Only 5% fully trust the automated evaluations that would enable such decisions.
This gap—autonomy increasing faster than assurance—reflects a broader trend. Enterprises deploy agents first and develop control layers afterward. The next year will be a retrofit cycle, with budgets shifting to systems that make agents governable and dependable.