Designing Pricing for Agents

AI agents don't fit neatly into existing pricing frameworks. They operate autonomously, consume resources unpredictably, and deliver value in non-linear ways.

The agent pricing problem

How do you price software that thinks, decides, and acts on its own? Per-seat doesn't work. Pure consumption can feel unpredictable to buyers. The answer is likely hybrid.

Emerging models

• Per-task pricing: Charge for completed workflows or outcomes
• Tiered autonomy: Different price points for different levels of agent independence
• Value-based metering: Bill based on business impact, not compute consumed

Design principles

Make pricing predictable for buyers and sustainable for builders. Offer caps, budgets, and transparent usage dashboards. Agents may be autonomous — but pricing shouldn't feel out of control.