Non-human identities now outnumber human users by anywhere from 25 to 50 times in the typical enterprise, according to an analysis published by InformationWeek. The report examines how AI agents, service accounts, and automated bots have rendered traditional per-seat software pricing obsolete, creating a structural mismatch between what organizations pay for and what actually drives value inside their platforms. As these non-human workers handle real tasks and consume actual resources around the clock, the decades-old model of charging by head count has begun to collapse.
The report identifies two common failure modes that emerge when per-seat pricing meets AI-driven workforces. The first is shelfware: licenses remain unused because team members access the platform only sporadically, yet organizations continue paying for potential rather than actual usage. The second is access restriction, where teams deliberately limit who receives a license to control costs, effectively capping the platform's value to make the budget math work. Both traps stem from pricing models disconnected from the work being performed, the analysis finds. Meanwhile, most enterprises continue budgeting for software based on head count and anticipated employee growth, even as agents have become the fastest-growing group in most environments.
According to the report, this disconnect creates activity that drives genuine value inside platforms but falls entirely outside what pricing models were built to measure. The analysis states that these systems perform meaningful work and require governance, yet none holds a seat license. The report frames the fundamental problem simply: when software is priced by head count and the workforce increasingly lacks a head, a structural gap emerges between expenditure and benefit. If enterprises don't make the shift to usage-based pricing, the current model doesn't merely become inaccurate—it actively undermines AI adoption, the report warns, with every new automation carrying a licensing cost that procurement can't trace to value creation.
The shift to usage-based pricing changes three things for buyers, the report explains. First, it creates organizational alignment by surfacing which teams drive usage, where value gets created, and who should own the AI consumption budget—a question seat-based models obscure. Second, it enables governance by making visible what ran, when, at what volume, and against which policies, allowing organizations to monitor AI workers the way they monitor human ones. Third, it establishes economic alignment: when agents become primary users, platform value no longer reflects how many employees have access but what those agents accomplish. The report draws a historical parallel, noting that cloud infrastructure adopted consumption pricing two decades ago, databases followed, and API platforms charge per call. Enterprise software has been the laggard in this pattern, but that's changing as AI fundamentally reshapes the relationship between people and work.
Organizations that adapt will stop measuring software success by login counts and start measuring by outcomes, the report concludes. Companies still optimizing for licensing efficiency rather than business outcomes will find their incentives running backward, penalizing the automation that should be driving competitive advantage. The analysis predicts that meaningful work will increasingly be performed by automations that neither collect paychecks nor occupy seats, making the core question not how many people log in but what gets accomplished. The pricing models enterprises choose today will either accelerate that transformation or quietly strangle it at the procurement stage. For buyers, the strategic risk isn't technical—it's whether legacy contract structures will let the AI workforce they're building actually do its job, or whether every efficiency gain will trigger a budget fight because the math was designed for a different era.