The Great SaaS Reset: Outcome-Based and Hybrid AI Agent Pricing in 2026

The Great SaaS Reset: Outcome-Based and Hybrid AI Agent Pricing in 2026

The traditional seat-based software-as-a-service (SaaS) business model is undergoing a fundamental restructuring in 2026. As autonomous AI agents perform more work historically handled by human employees, charging "per-seat" is proving non-viable.¹ Instead, the market is shifting toward outcome-based and hybrid credit models, led by major customer relationship management (CRM) and helpdesk platforms.

However, this transition is highly complex, marked by customer "metering anxiety," cost predictability concerns, and the technical challenge of verifying when a task is truly "resolved."

1. Zendesk Relate 2026: Double-Verified Resolutions

At its Relate 2026 conference (May 19, 2026), Zendesk made its bid to disrupt the SaaS landscape by expanding its Outcome-Based Pricing model. Under this model, Zendesk charges starting at $1.50 per resolution—completely moving away from traditional deflection-based metrics (which prioritized "containment" or blocking customers from human support) to verifiably completed outcomes.

• The Double-Verification Process: To overcome the "verification bottleneck" (where agents work faster than humans can audit them), Zendesk introduced a two-step check: the active AI agent first confirms that it has resolved the interaction, and then an independent AI evaluation model checks its homework. Spam and routine exchanges are automatically excluded.
• Commercial Targets: Zendesk is targeting $500 million in AI annual recurring revenue (ARR) in 2026, positioning its "Resolution Platform" as an autonomous workforce layer.

2. Salesforce Agentforce: The Whiplash from Conversations to "Flex Credits"

Salesforce's monetization of its "Einstein Copilot" rebrand, Agentforce, has served as a primary case study in the volatility of AI pricing.

• The 2024 Failure of "Per-Conversation" Pricing: At launch in late 2024, Salesforce priced Agentforce at $2 per conversation. This met with severe customer backlash and "sticker shock." CFOs feared a "blank check" scenario with no cost predictability, and the definition of a "conversation" was highly ambiguous. By May 2025, only ~8,000 of Salesforce's 150,000+ customers had leveraged Agentforce (single-digit adoption).
• The May 2025 Pivot to Flex Credits: Salesforce overhauled its model on May 15, 2025, introducing Flex Credits priced at $0.10 per "action" (sold in packs of 100,000 for $500, with one action consuming 20 credits). This aligned costs with discrete AI tasks (e.g., updating a record, running a workflow) rather than abstract conversations.
• The Hybrid "Flex Agreement" (Agentic ELA): To ease enterprise budgeting fears, Salesforce introduced the ability to convert unused seat licenses into Flex Credits and vice versa.
• Unlimited Per-User Ceilings: By mid-2026, Salesforce has fully embraced a hybrid menu. For internal use, it offers flat per-user-per-month add-ons (e.g., $125/user/month for Sales/Service Cloud) or premium "Agentforce 1" editions at ~$550/user/month that bundle unlimited employee-facing AI usage with a pool of credits. This "ceiling" has dramatically eased metering anxiety for CIOs.

3. HubSpot Breeze: Switching to Resolved Conversations

In April 2026, HubSpot abandoned flat per-use fees for its Breeze AI Customer Agent, shifting from $1.00 per conversation to an outcome-based model of $0.50 per resolved conversation (equivalent to 50 credits, with credits priced at $10 per 1,000).

• Definition of "Resolved": HubSpot defines a resolution as a conversation where the agent shares a knowledge source or completes an action with no human handoff within 72 hours, or where a sales lead is marked qualified/partially qualified. Unresolved conversations cost nothing.
• The Credit Pool Trap: HubSpot bundles a small number of monthly credits in its Professional (3,000 credits) and Enterprise (5,000 credits) Hubs. However, high-volume users face significant costs (e.g., handling 16,000 conversations/month can add over $5,100/month in extra credit costs), which automatically upgrade without manual approval.

4. The Competitive Landscape & Market Pressures

• Intercom Fin: Remains a major competitor charging $0.99 per resolution, allowing customers to set strict usage caps and alerts to manage costs.
• Atlassian Rovo: Has shifted away from flat-fee AI, metering its Confluence/Jira AI assistant (Rovo) via pooled credits with overage billing.
• Google Vertex AI: Offers raw agent queries at a highly competitive rate of ~$0.012 per query, highlighting the rapid deflation of underlying LLM inference costs.

Summary of AI Agent Pricing Models (Mid-2026)