Zendesk Shifts AI Agent Pricing to Pay-Per-Resolution Model

Zendesk ties CRM AI cost directly to resolved tickets

Zendesk introduced outcome-based pricing for AI agents in August 2024, charging customers only when AI autonomously resolves a support interaction. The model eliminates payment for failed attempts, escalations, or human handoffs. For enterprise buyers evaluating CRM platforms in 2025, this pricing structure creates a measurable alternative to the seat-plus-consumption model that dominates the market.

Salesforce Service Cloud with Einstein and Agentforce, ServiceNow Customer Service Management, and HubSpot Service Hub all price AI as seat licenses combined with usage credits or capacity tiers. Zendesk's approach removes the up-front capacity commitment and directly links spend to deflection rate and cost per resolved ticket—KPIs already tracked by most support organizations. Futurum Group research director Keith Kirkpatrick notes the vendor "plans to charge for successfully resolved interactions," positioning it as the reference design for AI agent economics going into 2026.

Budget and risk implications shift to the vendor

Outcome-based pricing changes the risk profile of AI agent deployments. Enterprises no longer pay for experiments that fail to deflect tickets or for agents that require frequent human intervention. Cost scales with proven automation performance rather than estimated seat counts or token consumption. This simplifies business case approval: finance teams can model AI spend as a variable cost tied to ticket volume and resolution success rate.

The model also introduces new contractual risks. Buyers must negotiate transparent resolution criteria upfront—what qualifies as "resolved" versus "escalated" or "partially handled." Without clear definitions, disputes over billable interactions become inevitable. Enterprises also need robust QA and escalation frameworks to prevent agents from over-optimizing for resolution metrics at the expense of customer satisfaction or regulatory compliance. An agent that closes tickets by routing customers to FAQ pages may meet the billing threshold but fail the experience test.

Competitive pressure on seat-based CRM vendors

For high-volume B2C support operations in retail, telecommunications, and travel, outcome-based pricing reduces the financial barrier to piloting AI agents. Traditional CRM vendors face a clear decision: match the model or defend capacity-based pricing with superior accuracy, integration depth, or compliance features. Salesforce recently launched Agent Builder, a low-code tool for customizing AI agents within its platform, but pricing remains tied to Einstein usage credits and seat licenses. ServiceNow introduced ServiceNow Studio for building agentic customer service applications, also anchored to premium Now Assist and Pro+ SKU tiers.

Zendesk and Workhuman—which ties pricing to employee retention outcomes rather than per-interaction metrics—represent the early edge of value-based AI pricing. Workhuman's model applies outcome guarantees to HR workflows, not CRM, but both vendors are testing whether enterprise buyers will accept vendors sharing downside risk in exchange for alignment on measurable business results.

What to watch

Track whether Salesforce, ServiceNow, or Microsoft Dynamics 365 respond with outcome-tied pricing options in the next two quarters. Enterprises running competitive CRM evaluations should model total cost of ownership under both pricing structures—outcome-based and seat-based—using actual ticket volume, deflection targets, and resolution accuracy from pilot data. Negotiate explicit resolution definitions, audit rights, and billing dispute escalation paths before signing. If your support operation handles more than 100,000 tickets per month, request a cost comparison from Zendesk and your incumbent vendor using trailing twelve-month data. The math will show whether outcome-based pricing delivers a structural advantage or simply shifts the invoice line item.