The Mechanics of Destruction: How Agents Kill SaaS

The Mechanics of Destruction: How Agents Kill SaaS

The traditional SaaS model depends on seat-based pricing: the more employees you have using the software, the more you pay. This worked when humans were the only entities capable of operating business software. AI agents shatter this assumption entirely.

One AI agent can replace 10-15 human seats in most SaaS applications [1]. Unlike humans, agents don't need graphical interfaces—they interact directly with APIs or manipulate screens programmatically. They can process thousands of tasks simultaneously, work across multiple systems, and never take breaks or require onboarding.

Consider Anthropic's Claude Cowork, which automates legal document drafting, compliance checks, and administrative workflows by directly manipulating Word, Excel, and web applications [1]. A law firm that previously needed 20 Salesforce seats for case management might now need just two seats for human oversight while Claude handles the actual data entry, follow-ups, and reporting.

The pricing model breaks down completely when your "users" are autonomous agents that can complete tasks faster and more accurately than human teams. As venture capitalist Tom Tunguz observed, per-seat pricing creates a fundamental disconnect when agents operate independently of human intervention [1].

This shift is accelerating because AI agents can build custom solutions faster than procurement teams can evaluate SaaS vendors. Companies are discovering they can deploy internal AI systems for surveys, business intelligence, CRM, and project management in days rather than months [1]. The switching costs that once protected SaaS companies have evaporated.

The Carnage: Winners, Losers, and the New Hierarchy

The market has been ruthlessly efficient in separating survivors from casualties. Software companies with thin user interfaces and limited data moats are being obliterated, while those with proprietary datasets and deep workflow integration are holding ground.

The casualties are predictable: point solutions with simple workflows and high per-seat costs. HubSpot, Atlassian, and similar companies built businesses on organizing and automating tasks that AI agents now handle natively. The iShares Expanded Tech-SaaS ETF (IGV) is down 21% year-to-date and 30% from its September 2025 peak [1]. Software P/E multiples have collapsed from 84x to 22.7x as investors reprice growth assumptions [1].

The survivors share common characteristics: proprietary data, complex workflows, and governance requirements. Palantir is up 22% this year because its value lies in data integration and analysis capabilities that become more valuable when AI agents need to make sense of complex information [1]. MongoDB and other infrastructure companies are benefiting as AI workloads require more sophisticated data handling.

The adapters are pivoting fast. Salesforce launched Agentforce to position itself as an agent orchestration platform rather than just a CRM. ServiceNow is betting on Now Assist to become the backbone for enterprise AI workflows. These companies understand that their future lies in enabling agents, not replacing them with human interfaces.

Even creative software isn't immune. Adobe has lost $120 billion in valuation as AI agents handle routine design tasks, content generation, and workflow automation [1]. The companies thriving in this environment are those that enhance AI capabilities rather than compete with them.

The New Software Stack: From Seats to Outcomes

The collapse of seat-based pricing is forcing a complete reimagining of how software creates and captures value. Three new pricing models are emerging: outcome-based pricing (pay per result), usage-based pricing (pay per task or API call), and hybrid models that combine human seats with agent tasks [1].

Outcome-based pricing represents the most radical shift. Instead of paying for software licenses, companies pay for completed work: tickets resolved, documents processed, leads qualified, or reports generated. This aligns perfectly with AI agents' capabilities and eliminates the disconnect between software costs and business value.

Usage-based pricing treats AI agents like cloud infrastructure—you pay for what you consume. This model works well for companies that want predictable costs while scaling agent capabilities up or down based on demand. It also creates natural incentives for efficiency improvements.

The most successful new software companies are being built around these principles from day one. They design for agent interaction first, human oversight second. Their APIs are optimized for programmatic access. Their pricing scales with business outcomes rather than human headcount.

This shift demands new metrics. Monthly Active Users (MAUs) become irrelevant when your primary users are agents. The new benchmark is Autonomous Task Completion (ATC): how many business processes can the software handle without human intervention? Companies optimizing for ATC are building the infrastructure for the post-SaaS world.

The Builder's Advantage: Why Internal AI Beats Vendor Software

The most telling sign of SaaS disruption isn't the stock crashes—it's the rapid adoption of internal AI development. Companies like Netlify and StackBlitz have replaced multiple SaaS subscriptions with custom AI agents built in-house [1]. This trend accelerates as AI development tools become more accessible and businesses realize they can build exactly what they need.

Internal AI agents offer three decisive advantages over traditional SaaS: perfect customization, zero licensing costs, and complete data control. A manufacturing company can build agents that understand its specific processes, terminology, and workflows without compromise. A financial services firm can ensure sensitive data never leaves its infrastructure while still automating complex compliance tasks.

The development speed is shocking. Teams are building functional AI agents in days using tools that would have required months of traditional software development. The barriers to custom software creation are collapsing just as the costs of SaaS subscriptions are becoming harder to justify.

This doesn't mean every company will build everything internally. The winning approach combines custom agents for core workflows with specialized AI services for complex tasks. A company might build internal agents for customer service and project management while using specialized AI for legal analysis or financial modeling.

The key insight is that software is becoming a byproduct of business logic rather than a separate purchase. When AI can generate the interfaces, workflows, and integrations you need on demand, the value shifts from owning software to directing its creation and operation.

Nordic Builders and the Post-Code Future

From our perspective at Up North AI, this transformation validates our core thesis: code is becoming free, but judgment isn't. The companies thriving in this environment aren't those with the best software—they're those with the clearest understanding of what problems need solving and how to direct AI agents to solve them.

Nordic companies have natural advantages in this transition. The region's emphasis on pragmatic engineering, data privacy, and sustainable business models aligns perfectly with the requirements of agent-driven software. Nordic firms are less encumbered by legacy SaaS investments and more willing to experiment with new approaches.

The opportunity lies in building judgment-amplified tools for specific verticals. Instead of creating general-purpose SaaS platforms, Nordic builders can create AI agents that understand the nuances of shipping logistics, renewable energy management, or financial services compliance. The value isn't in the software itself—it's in the domain expertise embedded in the agents' decision-making.

Data-rich industries offer the strongest defensive positions. Companies with proprietary datasets, complex regulatory requirements, or deep workflow integration can build AI agents that competitors cannot easily replicate. This creates sustainable advantages in a world where basic software functionality becomes commoditized.

The transformation also demands new organizational capabilities. Companies need AI orchestration skills, not just AI implementation. They need teams that can design agent workflows, establish governance frameworks, and continuously improve autonomous processes. These are fundamentally different skills from traditional software procurement and management.

The Judgment Layer: What Survives When Software Becomes Fluid

As we watch $2 trillion in software value evaporate, the deeper question emerges: what creates lasting value when AI can generate any software on demand? The answer lies in the judgment layer—the human insight that determines what should be built, how it should behave, and when it should adapt.

The most valuable companies in the post-SaaS world will be those that excel at directing AI agents rather than building static software. They'll understand their business processes deeply enough to specify exactly what outcomes they want. They'll have the governance frameworks to ensure agents operate safely and effectively. Most importantly, they'll have the judgment to know when to intervene and when to let agents operate autonomously.

This shift represents the culmination of a trend we've been tracking: software is becoming a real-time expression of business intent rather than a fixed product. When AI agents can modify workflows, create new interfaces, and integrate systems dynamically, the competitive advantage shifts to those who can articulate their needs most precisely and adapt most quickly.

The $2 trillion software crash isn't just a market correction—it's the sound of an industry realizing that the future belongs to those who can think clearly about what they want to accomplish, not those who can navigate complex software interfaces. In a world where code is free, judgment becomes the ultimate competitive advantage.

Sources

1. https://tech-insider.org/saas-stock-crash-ai-agents-2-trillion-2026
2. https://www.oliverwyman.com/our-expertise/insights/2026/apr/how-agentic-ai-reshaping-saas-valuations.html
3. https://www.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2026/saas-ai-agents.html
4. https://intellectia.ai/blog/will-ai-disrupt-saas-business-model-2026
5. https://vikinggrowth.com/news/what-does-ai-mean-for-saas-valuations-in-2026
6. https://markets.financialcontent.com/stocks/article/marketminute-2026-2-24-the-1-trillion-software-carnage-how-ai-agents-broke-the-saas-model
7. https://www.theregister.com/2026/02/04/ai_replace_saas
8. https://www.linkedin.com/posts/amirashkenazi_my-prediction-for-2026-by-december-ai-activity-7414700384432050187-wkju