Silicon Valley is rallying around a new extinction narrative. Agentic AI, autonomous systems capable of executing workflows on their own, could make traditional software-as-a-service (SaaS) applications obsolete. Big Tech investors worldwide argue that if artificial intelligence agents can update customer relationship management (CRM) records, create project tickets, and resolve support requests autonomously, companies may soon question whether to continue to pay per-seat subscription fees for software designed primarily for human operators.
Public markets have reacted as if that future is already underway. Since early 2026 (January to February), the S&P 500 Software and Services Index has fallen roughly 30%, wiping out nearly $1 trillion to $2 trillion in market value amid fears driven by agentic AI.
The sell-off hit many of the SaaS pioneers, including Salesforce, ServiceNow, and Snowflake. The iShares Expanded Tech-Software exchange-traded fund (ETF) dropped more than 20% as investors began pricing in what some industry experts now call the “SaaSpocalypse.”
The launch of Anthropic’s Claude Cowork in late January amplified investor fears. The agentic AI platform introduced a “computer-using agent” (CUA) capable of autonomously operating desktop software across multiple applications, allowing it to complete complex workflows without requiring human interaction with traditional interfaces. In early enterprise tests, Claude Cowork demonstrated the ability to handle tasks typically performed inside specialized SaaS tools.
Inside SAP, one of the largest enterprise software companies in the world, however, the narrative looks very different. Christian Klein, SAP’s CEO, believes the industry is misreading the moment. He says AI agents will not eliminate enterprise applications. They will make them more important than ever.
“What we are seeing is the market over-rotating on the belief that AI agents will replace every application and every seat license overnight. And that’s simply not how enterprise technology works,” Klein tells Fast Company in an exclusive conversation. “The breakthroughs are real, no question. But the sell-off conflates the disruption of lightweight, stand-alone tools with the disruption of deeply integrated systems that actually run businesses.”
Klein, who joined SAP as an intern in the late 1990s, has witnessed multiple waves of disruption across enterprise software—from the dot-com collapse to the emergence of subscription pricing. He says each transition brought predictions that traditional enterprise software would disappear. Each time, software ultimately became more valuable.
From his perspective, agentic AI does not eliminate the need for systems that manage enterprise truth. “While building agents is becoming easier by the day, deploying them across end-to-end supply chains or financial close processes, with full compliance and audit trails, is much more complex. That’s where 90% of the effort has to be invested,” he notes.
Why the SaaS Boom Is Facing Investor Doubt
The SaaS model, built on recurring subscriptions tied to employee head count, rests on a simple assumption: As companies grow, the number of software seats they purchase grows with them. Agentic AI challenges that premise directly. Instead of logging in to a dozen enterprise applications and assigning work to multiple teams, a single employee can prompt an autonomous agent to complete a task. The seat-based subscription model that powered the SaaS boom over the past two decades is beginning to look fragile in a world where software no longer depends on human seats to generate value.
Early enterprise deployments are already showing signs of that shift. Some companies report declining seat demand in categories such as project management, CRM administration, and customer support workflows—areas where automation can quickly replace repetitive human tasks. Atlassian shares plunged roughly 35% earlier this year after the company reported its first meaningful decline in enterprise seat growth. Investor anxiety that AI agents could bypass traditional workflow interfaces also sent ServiceNow stock down about 11%—even after the company reported earnings that beat market expectations—triggering broader panic across the enterprise software sector.
Meanwhile, a new generation of venture-backed startups such as Adept.ai and Replicate.dev are building what some investors call “service-as-software” or “SaaS 2.0” companies. Instead of selling application licenses, these firms deploy autonomous agents that complete entire business tasks and charge customers based on outcomes rather than user seats.
Klein, however, argues that most discussions about the “death of SaaS” overlook a crucial distinction. Agents still need systems of record. Even the most advanced AI agents depend on structured business data, governance policies, and access controls to operate safely.
“An agent that doesn’t understand how your business actually runs will quietly cascade errors into wrong decisions and real financial losses,” Klein says. However, he adds that as agents take on workloads, pricing will evolve toward usage-based or outcome-based models. “This shift will favor platforms with deep workflow integration over solutions that sell licenses to individual users. The value will move to whoever owns the business context, the data, and the governance that make that agent reliable.”
According to the research and advisory firm Forrester, global SaaS spending is projected to grow from $318 billion in 2025 to $576 billion by 2029, a trajectory that suggests the enterprise software core is not disappearing. SAP’s recent financial results offer some support for that thesis. The company reported 30% growth in total cloud backlog in fiscal year 2025, reaching €77 billion ($88.7 billion), while cloud revenue rose 23% during the same period. Its cloud enterprise resource planning (ERP) suite grew nearly 28%. Perhaps more tellingly, SAP Business AI was included in about two-thirds of its fourth-quarter cloud deals.
The Future of AI Agents Lies Inside Enterprise Software
SAP has largely weathered the SaaSpocalypse, with its stock falling only modestly (about 13% through mid-February 2026), while maintaining a market capitalization of roughly $235 billion as of March. That performance has significantly outpaced many pure-play SaaS peers, including ServiceNow. Klein says that of SAP’s 50 largest deals in Q4, about 90% included either AI capabilities or Business Data Cloud, the company’s unified data layer.
“Customers want AI deeply embedded in the systems their operations already depend on, powered by data they can actually trust,” he notes.
The company’s flagship platform is Joule, a generative AI copilot and orchestration layer integrated across SAP’s enterprise software suite, along with a growing ecosystem of autonomous “Joule agents” designed to automate complex workflows for finance, procurement, supply chain management, and human resources. Instead of navigating software interfaces externally, these agents rely on the company’s existing business logic, process expertise, and data models.
SAP’s internal data infrastructure also plays a key role. SAP Knowledge Graph organizes relationships across enterprise data, enabling AI systems to reason about how business processes connect across departments. The goal, Klein explains, is to transform enterprise systems with AI into what he calls “operating systems for autonomous work.”
The company’s current hybrid model, combining RISE subscriptions with embedded AI orchestration, positions it as a platform adapting to the emerging service-as-software era. However, industry experts note that SAP software can still be complex to implement, and many customers remain in the middle of lengthy migrations from legacy on-premise systems to the company’s cloud platforms.
A Competitive Reset Across Enterprise Software
SAP is not alone in making this bet. Major cloud platforms are racing to embed agentic AI directly into enterprise data environments. Databricks, Google BigQuery, AWS Redshift, and Microsoft’s enterprise software stack are all integrating autonomous agents that operate within governed data layers. The competitive battleground has shifted from building better models to controlling enterprise context.
“The current AI moment is different in speed and scale. Whoever owns the business logic, the process orchestration, and the governed data layer wins. AI agents don’t float above enterprise systems. They need them as their operating foundation,” Klein says.
Some of the most vulnerable companies in the SaaS sector are narrow point-solution vendors. Applications that perform a single function, such as ticket management or basic analytics, face greater risk from automation. Dan Faulkner, CEO of SmartBear, says platforms that manage entire enterprise workflows appear more defensible.
“Many SaaS products have been accessible via API [application programming interface] for years. The agents aren’t bypassing the software; they’re just using the API to work with it, rather than the human-oriented GUI [graphical user interface] people use,” Faulkner says. “Enterprise software will undoubtedly have to adapt to accommodate the growing population of agentic versus human users. Agents will have different constraints and capabilities than humans, so we may start to think of agent-forward work streams and human-forward work streams.”
He explains that companies that fail to tap into the AI-oriented enterprise IT budget will struggle.
Likewise, Kate Leggett, VP and principal analyst at Forrester, says that enterprises often spend $10 million a year on CRM and $100 million a year on ERP to support business-critical operations. These systems are deeply integrated into their operations and designed to support workflows that adhere to country-dependent regulatory compliance standards.
“Companies are not risking regulatory penalties and ripping out these deep investments right now and substituting them for AI agents,” she says. “AI consumption-based pricing will increase over time. Some of these vendors offer flex credits to preserve overall spend for a customer, where seats can be converted to consumption credits.”
Leggett explains that while “vibe coding” and DIY software development are gaining attention, enterprises still need the expertise to build, maintain, and scale that code over time while ensuring it remains secure and compliant. For core business workflows, that remains far from a trivial task, she notes.
If SAP’s Klein is right, the future may look different. Instead of eliminating enterprise software, agentic AI could transform it into the infrastructure that governs autonomous work. The SaaSpocalypse may not signal the death of enterprise software. It may mark the beginning of its next evolution.
“Enterprise software isn’t just code. It’s knowledge of how the businesses worldwide actually operate—the processes, the rules, the edge cases, plus the hard-won trust that comes from keeping mission-critical processes running at scale. You can’t vibe-code that in a garage,” Klein says. “Speed matters, but depth wins.”
