Half of companies run AI agents. Only a quarter measure them.
A Liferay survey of 500 U.S. professionals, published in August 2026, found that 54% of companies are running AI agents in production or actively piloting them. The same survey found that only 25% measure AI's impact with clear KPIs, and only 24% have a company-wide AI usage policy.
The enterprise picture is even more aggressive. CrewAI's February 2026 survey of 500 senior leaders at companies with more than $100M in revenue reported that 65% already use agents and 100% plan to expand in 2026. Security and governance topped the list of evaluation factors at 34%.
Put those together and the gap is clear: adoption is a board-level priority, but most teams can't yet say whether their agents are working. My take for engineering leaders is to treat measurement as part of the build, not a follow-up. Before an agent ships, agree on three numbers: the task it completes, the cost per completed task, and the rate at which a human has to step in. Those three tell you more than any adoption headline.