The per-seat model is dying
This article was drafted with AI assistance and reviewed by our team before publishing.
If your SaaS product charges per seat, you're building on a foundation that's quietly cracking. Agentic AI — software agents that autonomously complete tasks across multiple tools and systems, without a human ever logging in or touching an interface — is severing the direct connection between user count and value delivered. That connection is the entire basis of per-seat pricing. And according to Gartner, the consequences are already pricing in. This isn't a distant disruption. It's a strategic decision you need to make in your architecture and pricing model right now, before you've scaled on a stack that becomes impossible to unwind.
$234 Billion Reasons to Pay Attention
In July 2026, Gartner estimated that $234 billion in enterprise SaaS spend is at risk from "agentic arbitrage" by 2030. Agentic arbitrage is the phenomenon where AI agents complete tasks inside your product — querying data, triggering workflows, generating outputs — without any human user doing it. No login. No seat. No billable activity under a traditional license model. The math breaks fast. If a company replaces 20 human seats in your CRM with an AI agent that does the same work through your API, and your pricing has no mechanism to capture that value, you've just lost that contract's revenue. They're still consuming your infrastructure. You're just not getting paid for it anymore. Gartner also predicts that 40% of enterprise applications will feature task-specific AI agents by 2026, up from less than 5% in 2025. That's not a gradual rollout. That's a step function change in how enterprise software gets used — and it's happening inside your existing customer base right now.
Why This Is Happening Now
The causes stack on top of each other, and they're structural. First: AI agents have become genuinely capable. Tools built on models from OpenAI, Anthropic, and others can now handle multi-step reasoning, context retention across sessions, and integration with external APIs well enough to replace human operators in defined workflows. What required a trained employee two years ago can now be delegated to an agent running on a schedule. Second: orchestration infrastructure is maturing. The Model Context Protocol (MCP) — an open standard that lets AI agents communicate with and control external applications — is being adopted rapidly. This means agents aren't just scraping data; they're operating your product from the outside in, using the same API surface your developers built for integrations. Your product becomes a tool that an agent picks up and uses. Deloitte's 2026 TMT Predictions frame this clearly: SaaS vendors are moving from being systems of record to becoming AI-consumable services. The vendors who don't architect for this shift don't disappear — they just get used without being compensated for it. Third: enterprise buyers are actively pursuing this. Procurement teams are explicitly looking for ways to consolidate software spend while maintaining or increasing operational throughput. An AI agent that replaces three SaaS tools with API calls is not a science project — it's a cost reduction the CFO is signing off on.
The Architecture Problem Nobody Wants to Talk About
Most SaaS products built between 2015 and 2023 were designed around user sessions. The mental model was: a person logs in, does a thing, logs out. The UI is the product. Authentication, billing triggers, usage metrics, audit logs — all of it was designed around that human touchpoint. When an agent operates your product via API, that model falls apart completely. You have no session. You have no UI interaction to log as a billable event. You can't count seats. If your metering infrastructure isn't built to capture API-level activity as a first-class event, you're blind to what's actually happening inside your own system. Deloitte's analysis of SaaS tollgating — the practice of creating pricing gates around high-value API access — identifies this as one of the defining commercial challenges of the next three years. The companies that solve it architecturally will retain revenue. The ones that try to patch it with legal terms and contract language will lose customers when renewal comes around. The solution isn't to block agents. Blocking API access is both technically porous and commercially self-defeating — it just makes your product less useful in an agentic world. The solution is to make your system agent-aware at the infrastructure level.
What to Actually Build Differently
There are three concrete areas where the architectural and commercial decisions need to happen simultaneously. **Move to outcome-based or consumption-based pricing.** This is the most urgent shift. Rather than charging per seat or per month flat, you need a pricing layer that captures value at the event level — API calls, records processed, workflows completed, tokens consumed. This doesn't mean eliminating subscriptions; it means adding metering that remains accurate even when no human is involved. Tools like Stripe Billing, Orb, and Lago are purpose-built for this. Implement them before you go to market, not after you've locked in enterprise contracts with flat pricing. **Build API-first, with MCP compatibility in mind.** Every capability your product offers through the UI should be accessible via a clean, versioned API. More specifically, you should be evaluating MCP layer compatibility now — this is becoming the default handshake between enterprise AI orchestration systems and external tools. If your API surface requires session-based auth, legacy SOAP integrations, or undocumented internal endpoints, you have architectural debt that agents will route around rather than through. **Instrument everything at the event level.** Event-driven backends — systems that log discrete actions as events rather than only tracking user sessions — give you the observability you need to both meter usage accurately and understand how agents are using your product. This isn't just a billing concern. It's a product intelligence concern. If 60% of your API traffic is now agent-driven, that changes your roadmap priorities entirely. You can't make that call if your analytics infrastructure only speaks in MAUs and session counts. The organizational implication is equally important. Pricing strategy, product architecture, and API design can no longer live in separate conversations. The founders and CTOs who treat this as a product-and-engineering decision alone — without involving pricing strategy from day one — will build the right technical infrastructure and still lose revenue because the commercial model doesn't capture the value the system delivers. This is the constraint that makes 2026 the year to act. The SaaS companies that redesign around agentic usage patterns now will have defensible pricing models and architecture when their enterprise customers arrive with agents in hand. The ones that wait will be renegotiating contracts from a position of weakness — or quietly watching revenue compress while usage climbs.