What is the customer paying for?

AI software can be sold by user, by usage, or through a combination of the two. The unit on the invoice matters because it determines which part of customer activity becomes revenue.

Two published examples

As accessed on 2 October 2026, Microsoft’s enterprise pricing page lists Microsoft 365 Copilot at US$30 per user per month, paid yearly, with a separate qualifying Microsoft 365 licence required. Salesforce’s Agentforce page offers consumption through Flex Credits or Conversations alongside per-user options. Prices, inclusions, and contract terms can change.

These offerings serve different workflows and are not a like-for-like price comparison. They demonstrate that per-user and consumption models coexist; neither establishes which approach earns the better margin.

A simple revenue sensitivity

Consider a hypothetical vendor charging $30 per seat per month. One hundred seats generate $3,000 of monthly revenue. If automation reduces the seat count to 80, an unchanged price produces $2,400. Keeping revenue at $3,000 would require $37.50 per remaining seat, 25% above the starting price. These are assumed inputs, not customer results or a forecast for either company.

Usage pricing changes the calculation. Revenue becomes billable volume multiplied by the realised price per unit. More automated work can increase that volume even if fewer people log in. But retries, failed tasks, and human review may add costs without adding an equivalent billable unit.

What to look for in results

Paid adoption, realised pricing, renewal rates, and cost per completed workflow are more informative than a feature launch. A useful model connects those measures to gross profit and distinguishes higher activity from better economics.

Research commentary. Examples explain a method; they are not investment recommendations. Sources were reviewed on 2 October 2026.