Salesforce has introduced a performance-based pricing model for its AI agent offering in Korea, charging customers only when the agent successfully resolves support cases. This represents a deliberate shift away from traditional seat-based or usage-based licensing, effectively placing resolution risk on the vendor rather than the buyer. The move signals confidence in Agentforce's capability to handle genuine support work, but it also raises a critical question: what constitutes "resolved" in Salesforce's framework, and how will this definition protect teams from disputes over partial resolutions, escalations that required agent intervention, or cases where the AI handled initial triage but humans closed the ticket?
The pricing model carries significant implications for CX leaders evaluating AI agent investments. For teams already running Agentforce, this creates a potential renegotiation opportunity—particularly in markets where Salesforce extends the model—but it also introduces measurement complexity. Support teams will need to establish clear resolution criteria upfront and monitor whether the agent's definition of "resolved" aligns with their own quality standards and customer satisfaction thresholds. The approach is commercially aggressive and suggests Salesforce believes its agents can achieve resolution rates high enough to remain profitable under this constraint, which either reflects genuine confidence in the technology or a calculated bet that most organisations will still require significant human oversight, thereby limiting the agent's actual resolution volume.
This pricing experiment also signals a broader industry shift toward outcome-based models in CX AI. Rather than paying for access to a tool, teams pay for results—a framework that could reshape vendor competition if it gains traction across regions. However, the success of this model depends entirely on implementation rigour: teams must define resolution metrics before deployment, establish clear handoff protocols for cases the agent cannot resolve, and maintain visibility into what the agent is actually closing versus what it's merely progressing. For smaller vendors without Salesforce's financial runway to absorb resolution failures, this pricing approach may remain inaccessible, potentially widening the competitive moat for enterprise platforms that can afford outcome-based risk.
Salesforce launches AI agent, charges only for resolved support cases in Korea - CHOSUNBIZ Chosunbiz