Fixify Finds AI Agents Handle 1 in 3 IT Actions as Human Approval Climbs to 41%
Updated
Updated · Computerworld · Aug 4
Fixify Finds AI Agents Handle 1 in 3 IT Actions as Human Approval Climbs to 41%
2 articles · Updated · Computerworld · Aug 4
Summary
Nearly 18,000 plans and 147,000 agent actions across 40 companies showed AI now executes about one-third of enterprise IT workflow actions, mainly in repeatable, low-risk software, security and collaboration tasks.
Human oversight is still central in higher-stakes work: approval of AI-proposed actions rose to 41% from 23% over three months, while rejection fell to 16% from 27% as feedback loops improved.
About 23% of agent recommendations diverged from human judgment, and nearly half of those failures were 'target not found' cases tied to stale or messy operational data rather than core AI model problems.
Identity-lifecycle work such as onboarding, offboarding and IAM failed three-to-nine times more often than hardware or connectivity changes, making clean identity data and reliable integrations a key constraint on wider automation.
Fixify said enterprises should redesign IT operations around supervision—review queues, playbooks and easy-to-read approval surfaces—while leaving consequential decisions and exception handling with human analysts.