Most of the companies furthest along with AI expect to employ more people, and that is the number everyone is sharing. In a Box survey of 1,640 enterprise IT leaders, 79% of the most AI-mature organizations expect total headcount to grow over the next three years, against 58% of organizations overall. Only 9% say AI agents are mainly eliminating roles today.
The easy reading is that AI creates jobs, so everyone can relax. That reading is not wrong, and it has history on its side. The internet grew marketing teams and cloud computing grew IT departments, at least for a while.
The same report holds a second number. Nearly all organizations (96%) say it is important or very important that AI agents can reach the company’s own content and knowledge, the files and documents the agents would work from. Yet only 36% of those using or testing agents have connected them to trusted internal content across many uses. Most organizations are hiring ahead of the foundation their agents need.
Hiring ahead of the foundation is an old habit. When in doubt, companies add people: more teams to run AI, more to govern it, more to feed it and fix it. Long before AI, the reflex worked the same way. A process broke, a system creaked or a report kept arriving late, and the answer was another hire. Throwing people at broken things is a core enterprise skill, older than the spreadsheet.
So AI arrives on top of the same problems. The foundation is missing, the data is not connected, the process is still a mess, and the old answer returns. Few companies asked whether this time might be different, or whether AI was meant to repair the foundation that the extra hires now cover for. The habit runs again, with a better slide deck.
A rising headcount therefore says little about whether AI is working. In some companies it counts the people hired to cover for a foundation that was never built. Before the next role is posted, the open question is whether AI created new value that needs those people, or new complexity that now needs managing.
