Why the “SaaS-pocalypse” story is gaining traction
The recent panic around software-as-a-service reflects a real shift, but not the simple collapse narrative implied by the term “SaaS-pocalypse.” Investors have been selling off SaaS stocks on the view that AI agents and new coding tools could make large parts of the subscription software model look overpriced and overly rigid. If internal teams can generate tailored tools more quickly, or delegate work to increasingly capable agents, then the logic of paying for countless per-seat licenses begins to look less secure. What has changed is not merely sentiment, but the perceived balance between buying standardized software and building software that fits the business more precisely.
That idea has gained force because it no longer sounds theoretical. Executives cited in the source describe a market where companies are already building internal tools faster than before, even if they are not yet abandoning major SaaS platforms outright. Bill Vass of Booz Allen goes furthest, arguing that agentic systems could disintermediate both SaaS and older ERP vendors by allowing software to be generated directly from business policies and operational needs. The comparison to the pre-ERP era is telling: AI is reviving the possibility that enterprise software could become more custom again, after decades of forcing businesses to adapt themselves to standardized systems.
Customization is back, but the old problems have not disappeared
That does not mean the case for SaaS has suddenly vanished. The strongest defense from vendors is that customers are often underestimating what sits behind the visible interface. A dashboard can be cloned quickly, as illustrated by the example of CNBC reporters recreating a Monday.com-style workflow tool in under an hour, but the front end has never been the whole product. The harder problems remain data governance, reliability, security, backend maintenance and the slow accumulation of operational complexity once more features, users and edge cases begin to pile up.
This is where the anti-SaaS argument becomes more nuanced. AI may have commoditized parts of software creation, especially the visible and interactive layers, without yet eliminating the value of the managed systems underneath. Alexey Korotich of Wrike makes precisely this point, arguing that building a workable prototype is easy, while sustaining a dependable platform is not a one-person exercise but an ongoing team effort. In that sense, SaaS is not being invalidated so much as being pressured to prove that what it sells is durable infrastructure rather than merely convenient packaging.
The business model is likely to change before the category disappears
A more plausible outcome than wholesale extinction is a restructuring of how SaaS is priced, sold and evaluated. Faisal Masud at HP suggests that the traditional seat-based subscription model is already evolving, with outcome-based pricing emerging as a serious possibility. That shift is important because it recognizes the pressure AI is placing on one of SaaS’s core assumptions: that value scales neatly with user counts and usage tiers. If AI reduces the number of human seats needed to complete work, then vendors will need stronger arguments for why customers should keep paying under the old formulas.
This pressure is arriving at a moment when enterprise buyers are already frustrated. The source notes that software price increases have been outpacing CIO budgets, with many renewals coming with steep hikes, even as enterprise software spending continues to rise. Much of that spending, however, may be consumed by inflationary price pressure and AI infrastructure rather than enthusiastic expansion of legacy SaaS commitments. The correction, then, may first appear in pricing power and contract structure rather than in sudden platform abandonment. Smaller point solutions look especially vulnerable, while larger integrated systems may have more room to adapt.
The next winners will be those that turn SaaS into agentic infrastructure
The most credible voices in the article do not argue that SaaS disappears; they argue that it mutates. Vass expects a mixed future in which some companies build their own agentic systems while others rely on vendors that successfully package those capabilities into more flexible offerings. Tamar Yehoshua of Atlassian similarly argues that software will remain necessary because organizations will still need systems for coordination, communication and work management, even if AI radically changes how those systems are built and used. The category is not dying because businesses no longer need software. It is being forced to evolve because the terms under which software is bought and built are changing fast.
That makes the real dividing line less about whether a company is “SaaS” and more about whether it can make itself indispensable in an AI-native environment. Tatyana Mamut’s test is blunt but useful: the companies most likely to survive into the next decade are the ones that launch genuinely capable AI agents now, in 2026, not the ones that rely on yesterday’s seat-based logic and assume inertia will protect them. The future may still include SaaS, but it will almost certainly be a version that is less about selling access to static applications and more about delivering adaptive, agent-driven systems that absorb complexity without forcing customers to rebuild everything themselves.
Source: Is it really the end of SaaS as we know it?
