Zendesk introduces AI pricing based on completed work

Zendesk announced an AI billing model at its San Francisco AI Summit that charges businesses for eligible support issues resolved or tasks completed, rather than by user count. The change forms part of its new Outcome AI Platform.

Customers establish an “Outcome Budget” under the plans. Zendesk said it evaluates the quality and completeness of each result, allowing customers to identify which work counts against that budget.

The company also announced tools intended to support interactions with outside AI assistants. Its forthcoming Open Service Directory will list official brand support channels for assistants to find, including those of businesses that are not Zendesk customers. Betterment, ElevenLabs and Telekom Payment are among the early participants.

Zendesk said brands determine what information to share through the directory and when to require human involvement. The directory itself neither accesses private customer information nor performs actions.

Another tool, Zendesk Bridge, is in early access and connects external agents and applications, including ChatGPT, with Zendesk systems. Administrators can define permissions for activities such as creating or updating tickets. Bridge checks actions against those permissions and company rules, seeking human approval when required. A monitoring page records requests and blocked actions and provides a kill switch.

Two additional tools are generally available. AI Agent Launchpad reviews previous customer conversations to suggest work that AI could handle, while Agentic Analytics allows managers to query their data using everyday language and examine whether changes reduce repeat contacts.

In a Zendesk survey, 52% of consumers said they were comfortable using personal AI assistants for everyday tasks. However, 84% said access to human help should always remain available.