AI Service Pricing Shifts to Pay-Per-Result; Salesforce, Bairong Yunchuang Debut Outcome-Based Platforms

The pricing model for AI services is shifting from charging based on token consumption to paying for results. Pay-per-result means settlement is based on the business problems actually solved by the AI and the quantifiable value created, rather than on token consumption or call volume.

Some companies have already put this into practice. In June 2025, software service provider Salesforce launched Agentforce Help Agent, which adopts a pay-per-resolution model. Charges are incurred only when the agent autonomously solves a problem from start to finish; no fee is generated if the customer gives negative feedback or requests a transfer to a human agent. In December 2025, Bairong Yunchuang released a Result Cloud Platform that supports three value exchange models: task-based pricing, position-based salary system, and value-creation sharing.

Currently, token billing is the mainstream method, providing an objective, auditable, and uncontroversial measurement basis. The transition from concept to large-scale implementation of pay-per-result still faces multiple obstacles: there is no consensus on the definition and measurement standards for results; billing boundaries are blurred, as business outcomes are often the product of multiple interacting factors and cannot be easily attributed to a single model call; and there is a mismatch in settlement cycles—token billing is instantaneous, while pay-per-result requires post-hoc verification of effectiveness.