“AI Agent Air-Raid Warning”—Software Market at a Crossroads of ‘Collapse or Reordering’

Subscription Software Doomsday vs. Evolution Debate
Rise of Agentic AI Bypassing Individual Interfaces
Sophistication, Trust, and Infrastructure Cited as Survival Imperatives

The enterprise software market has reached a pivotal inflection point. Some artificial intelligence (AI) companies argue that a significant share of IT budgets will migrate to AI, raising the prospect of a sweeping market reconfiguration, while the industry remains divided between narratives of “collapse” and “evolution.” Proponents of the collapse thesis contend that AI agents will fundamentally alter how work is executed, destabilizing the traditional subscription model, whereas others maintain that incumbent services will absorb AI capabilities and reorganize into new forms. As a result, the market is likely to consolidate around firms that can demonstrate trustworthiness and operational competence.

Prospect of 50% Revenue Erosion

According to CNBC, Arthur Mensch, CEO of Mistral AI, said at the “AI Impact Summit” in New Delhi that “more than 50% of software-as-a-service (SaaS) spending currently purchased by corporate IT departments worldwide will soon shift to AI.” He added that “as AI dramatically accelerates software development speed, the conventional, standardized way of purchasing software is also changing.”

Mensch’s remarks extended beyond technological optimism surrounding AI firms to a forecast predicated on structural reallocation of enterprise IT spending, amplifying their impact. The comments coincided with recent weakness in global software stocks, heightening market unease. The iShares Expanded Tech-Software Sector ETF (IGV), which tracks major software names such as Microsoft and Salesforce, has declined, reflecting growing concern that the subscription-based revenue architecture underpinning SaaS could be disrupted.

The assumption is that AI will fill any vacuum left by SaaS. Mensch stated that “with the right foundational infrastructure in place, companies can connect their proprietary data to AI and build customized workflow systems in just a matter of days.” He further noted that “more than 100 corporate clients are seeking to dismantle outdated IT systems introduced in the early 2000s and rebuild them around efficient AI.” The implication is that enterprises are not merely maintaining or extending legacy systems, but actively redesigning platforms around AI-centric architectures.

The term “SaaSpocalypse” has emerged in the market to capture this anxiety. U.S. financial research firm Jefferies invoked the neologism while raising the possibility of the end of the SaaS era. During the first week of February, when Jefferies’ pessimistic outlook circulated, shares of Salesforce (-5.3%), Adobe (-5.66%), and ServiceNow (-9.68%) all fell sharply, erasing approximately $1 trillion in market value across the sector. The episode underscored that investors are reassessing not only individual company performance but the durability of the industry’s business model itself.

Counterarguments, however, remain substantial. Gartner projects that global IT spending will rise 10.8% year-on-year to $6.15 trillion this year. Although the growth rate of software spending is expected to ease from 15.2% to 14.7%, it still ranks second only to data center systems (31.7%) in expansion. Gartner observed that “it is unlikely that SaaS will be uniformly replaced in the short term,” while emphasizing that it is “highly probable that it will evolve and reorganize in conjunction with AI.”

Diverging Views on Agentic AI Commercialization

The emergence of agentic AI has lent weight to the collapse narrative, with Anthropic’s “Claude CoWork” at the center of the discussion. Released on January 30, the feature enables users to generate web pages, documents, and applications and execute 11 types of workflow automation through natural language dialogue, without programming expertise. Whereas traditional SaaS operates within predefined user interfaces and functional modules, CoWork can sequentially perform tasks such as file organization, spreadsheet classification, report drafting, and contract review within a user’s authorized scope. This ability to bypass function-specific subscription silos represents a significant departure from the structural premises of SaaS.

The technological shock intensified on the 5th of this month with the unveiling of “Claude Opus 4.6.” Anthropic demonstrated an experiment in which AI, without human intervention, wrote a C compiler capable of compiling the Linux 6.9 kernel. Sixteen Claude instances conducted 2,000 sessions over two weeks to complete a Rust-based compiler comprising approximately 100,000 lines of code, at a generation cost of $18,000. Opus 4.6 scored 80.9 on the SWE Bench Verified benchmark, a level comparable to a third-year mid-level engineer. Anthropic emphasized that tasks typically requiring months and substantial labor costs can now be completed in a short time at significantly lower expense.

Nevertheless, caution persists regarding real-world deployment. Gartner forecasts that by 2028 at least 15% of routine work decisions will be executed by autonomous AI agents, yet it also notes that large language models (LLMs) remain probabilistic systems at their core. Data bias or errors can be directly reflected in decision-making, and insufficiently designed access controls may introduce security vulnerabilities. Tom Coshow, a senior analyst at Gartner, remarked that “AI agents are tools to enhance SaaS, not technologies that replace it.”

Even so, pricing power among SaaS providers appears poised to weaken. An investment banking industry source observed that “enterprise solutions such as Salesforce’s customer relationship management (CRM) once possessed unassailable moats, but today AI agents are increasingly capable of replicating or substituting these functions at lower cost,” adding that “we are approaching a stage where AI not only delegates most human tasks but resolves issues autonomously.” According to this assessment, SaaS vendors in specialized enterprise or data service domains may face a dual challenge of declining market share and deteriorating profitability as AI agents gain traction.

Verifiability and Accountability as Critical Factors

Within the industry, there is growing consensus that generalized SaaS centered on broad functionality may recede, while firms equipped with industry specialization, advanced analytics, and regulatory compliance capabilities are more likely to endure. Services segmented by discrete functions—such as collaboration tools, CRM, or marketing automation—face direct overlap with the generalized execution capacity of agentic AI. In contrast, areas such as financial risk management, regulatory compliance, or long-term procurement systems, where legal accountability and audit trails are prerequisites, are considered less susceptible to immediate displacement. In these domains, data integrity, access control, and change tracking constitute core competitive advantages, and verifiability and clarity of responsibility are valued more highly than mere automation.

As AI absorbs repetitive tasks, the basis on which enterprises are willing to pay is also shifting. Previously, pricing was anchored in feature access rights and seat counts; now, integration capability, operational stability, and compliance readiness increasingly underpin pricing power. This evolution suggests that without parallel investment in control frameworks and governance architecture, the expansion of automation may amplify risk. Surviving firms are therefore expected to pivot toward commercializing data quality management and audit systems rather than simply expanding functional breadth.

Although it may be premature to declare the demise of the subscription model itself, industry consensus holds that seat-based billing alone will struggle to sustain growth trajectories. As agents traverse multiple applications to complete tasks, users demand integrated environments. Competitive differentiation will consequently hinge on application programming interface (API) stability, depth of system integration, and security certification frameworks. Observers increasingly characterize the SaaS market as entering a phase of realignment centered on trust and operational capability rather than pure feature competition.