Gartner Just Put a Price Tag on the SaaS Disruption: $234 Billion

The per-seat licence was the economic engine of enterprise software for thirty years. You bought a number of seats. Each seat was a user. Users logged into dashboards. Dashboards were the product. Gartner has now attached a number to how much of that model is at risk: $234 billion.

In an analysis first presented in its widely-cited “SaaSpocalypse” webinar series and now gaining sharp attention across enterprise software boardrooms as AI agents mature, Gartner estimates that $234 billion in enterprise application spending will be exposed to what it calls “agentic arbitrage” before the end of the decade. The term describes the mechanism by which autonomous AI agents, operating directly across backend systems without engaging the user interfaces that seat licences are built to monetise, strip the pricing rationale out of the dominant enterprise software model.

What the Analysis Says

Gartner’s SaaSpocalypse webinar frames the threat as structural rather than competitive. Traditional enterprise software disruption follows a familiar pattern: a better product wins users away from an incumbent. Agentic arbitrage is different. It does not require a better product to replace an existing one. It requires only that AI agents become capable enough to complete the tasks a piece of software was hired to do, by routing around the software’s interface entirely. An agent that can log into a CRM, pull contacts, draft personalised outreach, and schedule follow-ups through an email API has not replaced the CRM. The CRM still exists. But none of the activity that generated value for the software vendor — the human seat logged in, clicking through the dashboard — happened at all. The seat licence charged nothing. The vendor collected nothing.

Gartner separately predicts that by 2030, at least 40 percent of enterprise SaaS spend will shift toward usage-, agent-, or outcome-based pricing models as the seat-licence model breaks down under this pressure. That shift is not merely a pricing renegotiation between buyers and vendors; it is a wholesale restructuring of how enterprise software value is measured, captured, and contracted. Gartner also forecasts that 40 percent of enterprise applications will embed task-specific AI agents by the end of 2026, a figure that has more than tripled from under 5 percent in 2025 and reflects how quickly the structural conditions for agentic arbitrage are arriving.

The Mechanism: Why Seat Licences Cannot Survive Agents

The Dashboard Problem

The seat-licence model was built on a specific assumption about how software is used: a person sits in front of a screen, logs in, navigates a workflow, and produces an output. Pricing was attached to the human-in-the-loop moment — the seat — because that was the bottleneck. You could not do more work without adding more people, and you could not add more people without adding more licences. AI agents break that assumption at its foundation. An agent does not log in. It does not click through a dashboard. It calls APIs, reads and writes data, executes tasks, and produces outputs without engaging the user interface at all. If the software vendor’s pricing model only triggers when someone interacts with the interface, agents can extract the software’s underlying value while generating zero billable activity.

Which Categories Are Most Exposed

The exposure is not uniform across the SaaS market. Gartner’s analysis concentrates the risk in workflow-execution categories: HR and talent management software, CRM platforms, customer support tools, procurement systems, marketing automation, and any application whose primary value is delivered through a guided workflow rather than through proprietary data storage, industry-specific compliance infrastructure, or deep network effects. ERP systems like SAP and Oracle, and platforms like Salesforce with massive proprietary data moats, face less immediate displacement — their value is in the data and the ecosystem integrations, not the workflow execution layer. Agents will augment these platforms before displacing them. The categories facing urgent pressure are those built primarily on making human-driven task execution slightly easier, since those are precisely the tasks agents are being deployed to replace.

The Backstory: A Pricing Model Built for a Pre-Agent World

The $234 billion estimate lands in a context where the structural tensions have been building for several years but are only now reaching a point of acute enterprise attention. Deloitte’s SaaS sector analysis for 2026 noted that AI-native applications are beginning to compete not by being better software for humans but by being platforms that execute work autonomously — a fundamentally different value proposition. The growth of agentic AI on smartphones and the increasing willingness of enterprises to let agents act on their behalf in backend systems has brought that structural shift forward faster than most software vendors planned for.

The security implications of agents accessing enterprise systems directly are simultaneously reshaping the threat landscape. Our earlier reporting on agentjacking and AI coding agent vulnerabilities documented the emerging attack surface created when agents are granted the system-level access needed to execute tasks autonomously. That access is precisely what enables agentic arbitrage — agents need backend credentials to bypass the interface — and it is simultaneously what creates new categories of enterprise security exposure. The SaaS disruption and the SaaS security challenge are two sides of the same architectural shift.

Reactions

The response from enterprise software executives has, predictably, involved rapid repricing announcements. Microsoft’s Copilot and M365 agent pricing, Salesforce’s Agentforce licensing, and ServiceNow’s outcome-based contract experiments all reflect the same defensive move: get ahead of agentic arbitrage by restructuring pricing around agent actions or business outcomes before enterprise buyers do it unilaterally by refusing to renew seats. ServiceNow CEO Bill McDermott captured the vendor’s side of the calculation with notable bluntness earlier this year: agentic AI is “not just a revolution; it’s the only way to survive.” The subtext is that survival requires renegotiating the pricing model before customers renegotiate it for you by cancelling licences.

Enterprise buyers are arriving at the same conclusion from the other direction. Gartner’s research shows procurement criteria for software selection are already shifting: AI agent features have become a mandatory requirement, and vendors that cannot demonstrate how their product works alongside agents — not just with human users — are increasingly being filtered out at the selection stage. That represents a faster shift than most enterprise software procurement cycles normally allow.

The Dispute: SaaS Death or SaaS Transformation?

Gartner’s $234 billion figure has generated debate about what “exposed” actually means. Not all $234 billion will disappear from enterprise software budgets — some portion will be restructured into agent-based pricing, some will shift to AI-native competitors, and some will fund the build-your-own-agent workflows that large enterprises are increasingly deploying internally. The real question is how much of that spending stays with the incumbent SaaS vendors under new pricing models versus how much migrates to AI-native platforms or infrastructure providers. Gartner’s own scenario modelling includes a path where agentic AI drives 30 percent of enterprise application software revenue by 2035, suggesting the research firm’s view is not that SaaS collapses but that it transforms — with the vendors that adapt earliest capturing most of the restructured value.

What Happens Next

The practical timeline for enterprise SaaS buyers is more immediate than 2030. Gartner’s separate finding that over 40 percent of agentic AI projects will be cancelled by the end of 2027, due to escalating costs and unclear ROI, creates a cautionary counterweight. Buying fewer seats and deploying more agents does not automatically produce better business outcomes; it shifts the failure mode from under-adoption of software to poorly governed agent deployments. Enterprises that succeed will be those that treat the transition from seat licences to agent-based models as a governance and measurement problem as much as a technology replacement problem. The vendors that survive will be those that make their platforms legible to agents — clean APIs, explicit data contracts, programmatic access to workflows — rather than doubling down on the richly featured dashboards that agents are designed to ignore.

Why It Matters

The $234 billion figure is significant not because it represents a certain loss to enterprise software vendors but because it quantifies, for the first time with Gartner’s institutional weight behind it, the scale of the economic transition now underway. Enterprise software CFOs can no longer treat agentic AI as a feature discussion. It is a pricing discussion, a contract discussion, and increasingly a finance discussion — about whether the seat-licence renewals on the books represent real value that will be captured or theoretical value that agents will quietly route around before the next renewal cycle arrives.

Sources

Gartner SaaSpocalypse webinar: gartner.com/en/webinar/844664/. Gartner August 2025 press release on agentic AI enterprise applications. Gartner June 2025 predictions on agentic AI project cancellations. Deloitte 2026 SaaS sector analysis.