Human Seats and Metered Agent Actions

This pricing shape means AI software is becoming a blended labor model, not a pure software license. A customer may still pay for named humans who need dashboards and approvals, but also for the software workers doing the job, and for the discrete tasks those workers complete. That lets vendors keep the predictability of seat pricing while tying more of the bill to real work done, which matters once agents start replacing clicks, queue triage, and repetitive support or ops work.

• The closest live example is a hybrid. Salesforce now offers Agentforce with both per user licensing and action based Flex Credits, where each agent action draws credits from a usage pool. That is effectively human access plus metered agent work in one pricing system.
  1 sacra 4 salesforce
• Intercom shows why vendors add an action layer. Its support platform still charges per seat, but Fin AI Agent is billed at $0.99 per resolved conversation. The human pays for the workspace, the agent pays for successful work completed.
  2 sacra 5 intercom
• The operational challenge is billing complexity. Zapier has moved toward subscription plus pay as you go task pricing, and Ampersand describes the same tension. Buyers like bills they can predict, but AI creates variable costs because every agent run can trigger many reads, writes, and follow on actions.
  1 sacra 3 sacra 6 zapier

Over the next few years, application pricing is likely to settle into bundles that include a base platform fee, a set of human seats, and a metered pool for agent work. The winners will be vendors that make that bill legible enough for finance teams and aligned enough with outcomes that customers feel they are paying for work, not shelfware.