From seats to actions

Traditional software-as-a-service billing has been anchored to per-seat subscriptions: companies pay for each employee with access to a tool. As AI agents take on tasks autonomously, that link between human users and fees weakens, since an agent can perform work at any hour regardless of how many people are licensed.

Microsoft's Copilot Cowork, presented on the company's official Copilot blog, reflects this shift by positioning AI as a collaborative "coworker" rather than a passive assistant invoked by a seat holder. Salesforce, for its part, has introduced prepackaged Agentforce offerings for customer help, signaling a move toward charging for packaged AI capability rather than purely for platform access.

The industry has not fully switched

The change is real but incomplete. Per-seat licensing remains the dominant model for most software vendors, and outcome- or usage-based AI pricing is still emerging rather than established as the standard. The exact economics of how vendors will bill for agent work — per task, per conversation, or per outcome — have not been uniformly settled across the market.

The Salesforce newsroom tiles reviewed reference the company's fiscal results, but the claim's broader framing — that billing methods will comprehensively change in the AI era — extends beyond what any single product announcement demonstrates.

For enterprises, the transition carries practical implications: budgets may move from headcount-linked license counts to consumption metrics, complicating cost forecasting but potentially aligning software spending more directly with business output.

In the era when AI does the work, software billing is indeed changing toward service and usage-based models, though the full transition remains underway rather than complete — making the claim Mostly True.