New research from Fixify reveals a growing trend in enterprise IT: AI agents are handling roughly one-third of workflows, while human analysts focus on higher-stakes tasks and exceptions. The study analyzed nearly 18,000 plans and over 147,000 actions across 40 companies.
Rather than replacing IT professionals, AI is augmenting their capabilities. Agents excel at routine tasks like software updates, security group management, and application access changes—areas where requests are typically repeatable and have clear reversal paths. Meanwhile, human analysts manage complex issues, provide oversight, and shape how these systems operate.
As feedback loops improve, AI’s role continues to expand. Over the study period, human approval of agent-proposed actions increased from 23% to 41%, while rejection rates fell from 27% to 16%. This suggests that as agents learn and adapt through real-world interactions, they become more reliable.
The most common use of AI in IT involves ‘running skills’ (actual task executions) at 39.4%, followed by coordination activities like sending messages and providing updates (57.5% combined). While full workflow automation remains limited (1.1%), agents are increasingly handling entire identity-lifecycle processes—onboarding, offboarding, access management.
However, this is also where AI encounters the most challenges. Identity-related tasks fail 3-9 times more often than hardware or connectivity changes, primarily due to data inconsistencies and incomplete information. When an agent can’t find a user, group, or resource it needs (the ‘target not found’ error), it highlights underlying issues with IT hygiene rather than AI capability.
According to Fixify CEO Matt Peters, this human-in-the-loop approach creates a ‘scaffolding’ effect where AI handles the predictable parts while humans manage exceptions and ensure quality. Rather than seeking complete automation, organizations should focus on how AI can extend human capabilities—pointing ‘conversational intelligence’ in the right direction to solve specific problems.