Work-as-a-Service: The Post-SaaS Business Model
SaaS (Software-as-a-Service) sold access to tools — you pay per seat per month for the right to use software. Work-as-a-Service (WaaS) sells completed outcomes — you pay per resolved support ticket, per generated report, per closed deal. The shift from selling tool access to selling work completion is the most fundamental change in software business models since SaaS replaced perpetual licenses. It changes pricing (per-outcome vs per-seat), margins (variable AI compute cost vs near-zero marginal cost), retention metrics (outcome quality vs feature stickiness), and valuation frameworks (revenue per outcome vs revenue per seat).
The Death of the Seat
A Zendesk seat costs $89/month regardless of whether the agent resolves 10 tickets or 100. The company makes the same revenue either way. In Work-as-a-Service, you pay $1.50 per resolved ticket. If the AI agent resolves 1,000 tickets, the cost is $1,500 — and no human agent was needed.
This changes the economics completely. The customer pays less per unit of work. The provider earns more total revenue if they can scale volume. But the provider also bears the compute cost — every resolved ticket requires inference compute. This is the Agentic Accounting problem: AI inference becomes Cost of Goods Sold.
Three Pricing Models Emerging
Success Credits: Companies buy credits that are consumed when outcomes are delivered. Salesforce Agentforce uses this model — customers purchase "agent conversations" in blocks. The value is tied to resolution, not access.
Outcome-Based Pricing: Price is tied directly to measurable outcomes — $X per qualified lead, per resolved ticket, per generated report. This requires robust measurement infrastructure and creates a new category of "outcome verification" tools.
Consumption-Based (Hybrid): Base platform fee plus per-action charges. Similar to Snowflake's compute credits model. Predictable base revenue with upside from agent activity volume.
Who Wins the WaaS Transition
Winners are companies that can own the outcome measurement layer: if you can prove the agent resolved the ticket, generated the lead, or completed the analysis, you can charge per outcome. This favors verticals with clear success metrics over horizontal tools with ambiguous value.
Platforms that become the runtime for work — where agents execute, state is managed, and outcomes are tracked — capture the infrastructure tax. This connects to the Cloudflare Toll Booth thesis: whoever provides the agent execution layer collects revenue on every outcome.