The Per-Seat Model Meets Its Reckoning
For two decades, enterprise software has been sold by the seat. You counted your users, multiplied by a per-user fee, and that was the deal. Artificial intelligence agents are quietly demolishing that arithmetic, and the largest platform vendors have noticed. ServiceNow, SAP, and Workday are all drawing new lines around the customer data stored inside their systems, because a future in which a handful of AI agents do the work that once required thousands of licensed humans is a future in which per-seat revenue collapses. The vendors are moving to protect the franchise before it erodes.
The logic is uncomfortable but clear. If an enterprise can deploy a small number of autonomous agents to perform tasks across HR, finance, and IT that previously demanded armies of individual users each holding a paid seat, then the seat count, and therefore the vendor's revenue, falls even as the value delivered rises. The systems of record become more essential and more used, yet the pricing mechanism that monetized them stops tracking that usage. No incumbent is going to watch that happen passively, and none is.
Action Fabric and the New Toll Booths
ServiceNow's response is the most architecturally explicit. The company introduced Action Fabric, an integration layer that external AI agents must pass through to access data and execute workflows inside its platform. Framed benignly, it is a governance and security control, a single, monitored doorway through which outside agents enter rather than a sprawl of ad hoc integrations. Framed commercially, it is a toll booth. By making agents route through a controlled layer, ServiceNow gains both the ability to govern that access and the leverage to price it.
This dual character is the whole point. Enterprises genuinely do need a governed way to let third-party agents interact with sensitive systems of record, and an uncontrolled free-for-all of agents reading and writing HR or financial data would be a security nightmare. But the same chokepoint that provides governance also provides monetization, and the vendor that controls the doorway controls the terms of entry. ServiceNow, SAP, and Workday are each building their own version of that doorway, and each intends to charge for passage through it.
From Seats to Consumption
The pricing evolution underway is a shift from counting people to counting activity. Instead of licensing a fixed number of human users, vendors are moving toward models that meter what agents actually do: the transactions they execute, the data they consume, the workflows they trigger. This aligns revenue with value in a way per-seat pricing no longer can, because it captures the intensity of automated usage rather than the headcount of a shrinking human workforce. In theory it is a fairer model. In practice it introduces a new and less predictable cost dynamic.
That unpredictability is the rub for buyers. Per-seat pricing had the virtue of being legible and stable: you knew your user count and could budget accordingly. Consumption pricing tied to agent activity is far harder to forecast, because an enterprise scaling up its use of autonomous agents may not know in advance how many transactions those agents will generate. The very automation that makes agents attractive can make their costs balloon in ways that are difficult to see coming, and finance teams are only beginning to build the muscles to model it.
What CIOs Have to Do Now
For technology leaders, the practical consequence is that agent access to systems of record has become a first-class contractual and governance question. When ServiceNow, SAP, or Workday comes up for renewal, the negotiation can no longer stop at seat counts and support tiers. It has to address how external agents will be permitted to interact with the platform, what that access will cost, and how usage will be measured and capped. Leaders who treat these as afterthoughts risk discovering the terms only when an unexpected invoice arrives.
There is a strategic dimension beyond the line items. As vendors erect gated layers around their data, enterprises should press for portability and clarity: the ability to understand exactly what their agents are doing inside each platform, the ability to move workloads if terms sour, and transparency into how consumption is metered. The organizations that engage early, while the pricing models are still forming, will help shape terms they can live with. Those that wait will inherit whatever framework the vendors settle on without them.
A Preview of the Whole Software Economy
We see the maneuvering by these three vendors as an early and unusually visible instance of a transition every software company will eventually face. The per-seat model was built for a world where software augmented human labor and humans were the unit of value. Agentic AI breaks that assumption by substituting for labor rather than merely assisting it, and any pricing model anchored to human headcount will come under the same pressure that ServiceNow, SAP, and Workday are now responding to ahead of the curve.
How this resolves will shape enterprise IT budgets for years. If consumption pricing becomes the norm, the cost of enterprise software will grow more variable and more tightly coupled to the intensity of automation, rewarding careful governance and punishing sprawl. The vendors are betting that owning the systems of record, and the gated access to them, keeps them indispensable regardless of how the human seat count evolves. For their customers, the task is to make sure that indispensability does not translate into unbounded and unaccountable cost.
