The Death of Seats: How Agentic Pricing Models Are Rewriting Enterprise Software Economics

The $600 Million Question

ServiceNow's Now Assist surpassed $600 million in Annual Contract Value by the end of 2025: this single figure represents more than the total revenue of most public software companies. It signals the end of an era. After two decades of pricing software by the number of human users, enterprise vendors are discovering that artificial intelligence breaks every assumption about how value flows through organizations.

The shift is structural, not cyclical. When ServiceNow CEO Bill McDermott revealed that half of all new business bookings now come from pricing models that do not rely on seat-based approaches, he was describing the quiet dismantling of the economic foundation that powered companies like Salesforce, Zoom, and hundreds of other SaaS giants. The correlation between headcount and software spend—the bedrock principle that made recurring revenue predictable—is dissolving.

The Measurement Problem

Salesforce's introduction of "Agentic Work Units" (AWUs) shows both the promise and the challenge of this transition. The company reported 2.4 billion AWUs delivered across its platform, with 771 million in Q4 2025 alone: a 57% quarter-over-quarter increase. An AWU measures "one discrete task accomplished by an AI agent," from updating records to triggering workflows to resolving support cases.

This metric represents a fundamental philosophical shift: from charging for access to charging for output. But it also reveals the industry's struggle to define value in an agentic world. Critics note that AWUs measure machine exertion rather than business outcomes. An AI agent caught in an infinite reasoning loop generates AWUs with each failed attempt; customers effectively pay for the software's confusion.

The challenge mirrors what Ben Thompson described in his analysis of platform dynamics: "The hard part isn't building the technology; it's aligning economic incentives with actual value creation." When ServiceNow charges for "assist tokens" or Salesforce bills for "work units," they are attempting to solve a measurement problem that has no historical precedent.

The Consumption Trap

The migration toward usage-based pricing creates new volatilities that challenge the SaaS orthodoxy. Goldman Sachs research from February 2026 showed that 49% of institutional allocators planned to increase exposure to software—the highest net figure since 2017—but they are demanding transparency about consumption patterns rather than predictable per-seat growth.

This shift forces CFOs to rethink budgeting fundamentals. Microsoft Copilot costs $30 per user monthly, but only with an existing Microsoft 365 license; the true cost burden is significantly higher. Salesforce Agentforce operates on consumption-based billing tied to conversations and tokens. The result: 78% of IT leaders surveyed by Zylo reported unexpected charges due to consumption-based or AI pricing models.

"When you add AI and consumption-based pricing, we're talking about more budget volatility and pressure on in-year spend, which kills innovation," observed Jez Back, Cloud Economist at Capgemini Invent. Organizations incur surprise charges mid-contract as AI usage scales, creating friction between procurement teams and operational departments.

Platform Consolidation Accelerates

The pricing model disruption is accelerating platform consolidation. ServiceNow's positioning as an "AI Control Tower"—managing autonomous agents across entire technology stacks—reflects recognition that fragmented point solutions cannot survive the transition to consumption-based billing. The company's $7.8 billion acquisition of Armis, integrated into its automation workflows, represents the strategic imperative to own the orchestration layer.

Smaller SaaS providers without proprietary data layers are becoming acquisition targets. They lack the scale to implement complex AI billing models or the data moats necessary to train effective agents. As Marc Benioff noted during Salesforce's Q4 earnings call, the winners will be platforms that can "bring humans and agents together on one trusted platform."

Historical Precedent and Structural Shift

The current transformation echoes Clayton Christensen's framework for disruptive innovation, but with a crucial difference: the disruption comes from within incumbent platforms rather than external startups. ServiceNow and Salesforce are cannibalizing their own seat-based models before competitors can do it for them.

This parallels the utility industry's evolution from flat access fees to usage-based billing as value became more variable. The difference is speed and scale: software companies must execute this transition while maintaining growth rates that satisfy public market expectations. ServiceNow's guidance targeting $15.53-$15.57 billion in subscription revenue for 2026 suggests confidence that the new models can sustain growth momentum.

The Credit-First Architecture

Pricing experts identify "credit-first, agent-aware" architectures as the emerging standard. Customers purchase pools of AI capacity and allocate them across use cases, similar to cloud computing credits but tied to business outcomes rather than compute cycles. This approach addresses the volatility problem while maintaining alignment between usage and value.

The shift toward outcome-based pricing—where vendors charge per successful resolution, completed transaction, or verified business result—represents the logical endpoint of this evolution. Intercom's Fin AI agent charges $0.99 per successful customer resolution; if the AI fails, customers pay nothing. This model transfers risk to vendors and forces them to optimize for effectiveness rather than activity.

ServiceNow's march toward the $1 billion ACV milestone for Now Assist will determine whether the industry can successfully navigate this transition. Can enterprise software companies maintain their premium valuations while fundamentally restructuring how they capture value? The seat-based model created predictable recurring revenue; consumption-based models create volatile but potentially larger revenue pools tied directly to customer outcomes.

The answer will define not just the future of SaaS, but the economics of enterprise technology itself.