Enterprise software is quietly filling up with AI agents, and most IT departments still track their devices the old-fashioned way.
Gartner forecasts that task-specific AI agents will be built into 40% of enterprise applications by the end of this year, up from under 5% previously. Gartner doesn't specify when that lower figure was measured, but the direction is clear: agents are moving from novelty to default feature fast. Meanwhile, a survey from endpoint management vendor Automox found that only 46% of organizations currently automate the inventorying and monitoring of what's actually running on their managed devices. The rest are doing it manually, sporadically, or not at all.
That gap matters more for agents than for ordinary software. An unpatched app is a passive risk sitting on a disk. An AI agent takes actions - reading files, calling APIs, moving data - and if IT can't see it running, nobody's watching what it does with that access. Endpoint management has always lagged behind whatever's newest on the network; agents are just the latest thing arriving faster than the tooling built to catalog it.
This is the same shadow-IT problem that showed up with cloud apps and BYOD, just with software that now acts on its own instead of waiting to be clicked.