SaaS is already dead

Enterprise AI spending is on track to hit $2.52 trillion in 2026, a 44% jump from the year before. Anthropic's revenue grew over 428% last year. Cursor blew past $2 billion in annualized revenue by early 2026. Meanwhile, legacy SaaS companies are watching their stock prices slide, their growth stall, and their customers quietly cancel seats. This isn't a story about AI adding a new feature to an old product. It's about the business model underneath SaaS, the one built on per-seat pricing, bundled features, and switching costs, breaking apart. The disruption isn't on the horizon. For a growing number of software categories, it already happened.

The business model breaks first

The traditional SaaS playbook is elegant in its simplicity: build a workflow tool, charge per seat, expand within the org. Revenue scales with headcount. Switching costs compound over time. Investors love the predictability. But per-seat pricing has a fatal assumption baked in: humans are the users. When an AI agent can do the work of five seats, the math collapses. You're not losing a feature comparison, you're losing the unit of value. A company that used to pay for 50 seats on a project management tool now needs five humans and an agent layer. The SaaS vendor didn't get outbuilt. They got out-modeled. Forbes put it bluntly: agentic AI is breaking the SaaS pricing model. Not because the software is worse, but because the relationship between software usage and human headcount no longer holds. RSM's analysis reinforces this, arguing that the subscription-based model of charging a fixed monthly fee per user, the very thing that propelled software valuations for two decades, is about to break. SaaS companies are scrambling to respond. Some are moving to usage-based pricing. Others are experimenting with outcome-based models. According to Deloitte, 83% of AI-native SaaS companies already offer usage-based pricing. But for incumbents, the transition is painful: it risks disrupting existing customer relationships, confusing buyers, and compressing revenue in the short term.

AI-native tools are eating from below

The most telling signal isn't what's happening to legacy vendors. It's what's happening to the newcomers. Cursor, the AI-native code editor, surpassed $1 billion in annual recurring revenue in November 2025. By February 2026, its annualized run-rate exceeded $2 billion. That's order-of-magnitude growth within roughly a year. Lovable, a "vibe coding" startup that turns natural language into full-stack web apps, hit $400 million in ARR by March 2026, doubling from $200 million in about a year. It reached $100 million faster than OpenAI, faster than any software company in recorded history. These aren't niche tools. They represent a fundamentally different approach: instead of giving humans a better interface to manage workflows, they collapse the workflow entirely. You don't use a project management tool to track the development of a feature. You describe the feature and an agent builds it. The competitive dynamic is different too. These tools aren't going head-to-head with incumbents on feature checklists. They're making entire product categories feel unnecessary. That's not competition. That's displacement.

The great budget shift

The money tells the story. Gartner forecasts worldwide AI spending will total $2.52 trillion in 2026. A BCG survey of over 2,300 senior executives found that companies expect to spend approximately 1.7% of revenue on AI investments this year, more than double the 0.8% average in 2025. But here's the critical detail: most companies aren't expanding their total IT budgets to accommodate AI. They're reallocating from somewhere else. SaaStr laid out the math. Average IT budget increases sit around 2.79%. Average vendor price increases are at 9%. Companies are underwater before they even think about new purchases. So where does the AI budget come from? It gets stolen from existing SaaS line items. The TechRadar report on enterprise AI spend captures this shift vividly: AI is becoming the line item, not an add-on to existing software budgets. OpenAI and Anthropic are the big winners, absorbing a growing share of enterprise spend. Cursor saw budget allocation increases exceeding 600%. The redistribution is stark, concentrated among a small number of AI-native vendors while traditional SaaS providers lose relevance. This isn't a temporary reallocation. It's a structural shift in how enterprises think about software spending.

The API irony

Here's the part that stings: the SaaS companies that built the best integrations and APIs are actually the most vulnerable. For years, strong API ecosystems were a competitive moat. The more integrations a tool had, the stickier it became. But AI agents are exceptional integration layers. They don't need pre-built connectors. They can read documentation, authenticate with APIs, and chain actions across services autonomously. As one analysis from Uncover Alpha put it, most SaaS products have well-documented APIs, and AI excels as an integration layer without the need for pre-built connectors. The more open and clear a company made its API, the easier it became for agents to route around the product itself. The moat became the bridge. This is especially true for SMB customers. LinkedIn analysis of 24 publicly traded SaaS companies found that while enterprise customers with deep institutional complexity and compliance requirements remain sticky, SMB customers are increasingly replaceable by AI. Simpler workflows, less institutional complexity, lower switching costs. Monday.com and ZoomInfo both flagged this in their SMB customer bases.

What survives

Not everything dies. The Bain report on agentic AI and SaaS offers a useful framework: disruption is mandatory, but obsolescence is optional. The outcome depends on where a company sits along two axes, the potential for AI to automate user tasks, and the potential for AI to penetrate existing workflows. Companies that own critical data, manage compliance, or serve as systems of record have natural defensibility. Gartner predicts that 40% of enterprise apps will feature task-specific AI agents by 2026, up from under 5% in 2025. The survivors won't be tools that humans click through. They'll be platforms that AI agents use. The distinction matters. A CRM that requires a human to manually update deal stages is vulnerable. A CRM that serves as the data substrate for an agent that autonomously manages pipeline, sends follow-ups, and flags risk is something different entirely. Same category, different posture. AlixPartners frames this as a shift from selling tools that help humans work to selling digital workforces. The software companies that thrive won't be those counting the most seats, but those orchestrating the most efficient work. Some companies are already making this transition. Notion, for instance, has evolved from a note-taking app into a platform that combines CMS, database, and agent capabilities, positioning itself as infrastructure that agents operate on rather than a tool that agents replace. Whether that bet pays off long-term remains to be seen, but the strategic direction is clear.

The uncomfortable middle

The hardest part of this transition isn't the technology. It's the human cost. When a SaaS company loses 40% of its seats because an agent layer replaced most of the humans using the product, that's not just a revenue problem. It's a jobs problem. Customer success teams, implementation consultants, account managers, all built around a headcount-driven model that's unwinding. And it's not just the SaaS vendors. Their customers are going through the same reckoning. If an AI agent can handle the work that five people used to do through a SaaS tool, both the tool and the roles built around it face pressure. Deloitte's State of AI in the Enterprise report for 2026 notes that 90-95% of organizations are still seeing little to no measurable financial return from their AI investments. The disruption is real, but the transition is messy. Many companies are caught in an uncomfortable middle: old models are breaking, new models aren't proven, and the pressure to act is immense.

What this means going forward

The SaaS model isn't going to vanish overnight. Enterprise software is deeply embedded, contracts are long, and institutional inertia is real. But the direction is unmistakable. Three things are becoming clear:

The old playbook, build a feature, charge per seat, compound with switching costs, assumed a world where humans were the primary users of software. That assumption is no longer safe. The companies that recognize this and restructure around it will find enormous opportunity. The ones that don't will discover what it feels like to be on the wrong side of a platform shift.

References