Anthropic’s latest Artificial Intelligence (AI) launch has showed clear progress toward agentic behavior and bringing the vision of systems that don’t just assist humans, but increasingly act on their behalf, closer to reality. Although, within hours, US software stocks were sold off sharply and a familiar narrative resurfaced: AI is finally coming for Software-as-a-Service (SaaS).
The speed of the reaction was striking but is not completely surprising. Technology inflection points tend to trigger exaggerated conclusions, especially when they challenge long-held assumptions about how value is created in enterprise software.
Reach out to discuss this topic in depth.
Why everyone panicked
The panic was largely driven by what investors thought the launch implied. For years, AI had been framed as an assistive layer that is useful but contained. Anthropic’s release shifted that framing. If AI can reason, plan, and execute tasks autonomously, then a deeper question emerges: what happens to the application layer that sits between intent and execution?
Markets answered that question quickly and bluntly. If intelligence can act directly, software starts to look redundant. Years of SaaS economics were suddenly reinterpreted through a single lens: disintermediation.
What got lost in the rush was the difference between technical capability and enterprise operating reality.
For the past two years, enterprise software has already been going through a quieter reset. Growth rates normalized after the post-pandemic surge. Buying cycles lengthened. Chief Financial Officer (CFO) scrutiny increased. Enterprises began rationalizing bloated tech stacks and asking harder questions about utilization and Return on Investment (RoI).
This shift is already visible in the financial performance of large enterprise software providers. Many companies that were growing north of 20 percent in late 2021 are now growing at materially lower, but still healthy, rates. Microsoft’s revenue growth has moderated from the low twenties to the high teens. Salesforce has moved from roughly 20 percent growth to high single digits. ServiceNow and Workday remain comparatively stronger, but even there the trajectory reflects normalization rather than collapse.
In that context, Anthropic wasn’t the cause of the panic. It was the trigger. It compressed multiple slow-burn pressures into a single market moment and gave existing unease a focal point.
The assumption that broke the narrative
At the heart of the sell-off lies a simple assumption: that intelligence replaces software.
That logic might hold in low-friction consumer environments. It breaks down in the enterprise. platforms that are not just collections of features or dashboards. They encode how work actually flows through organizations along with how cases are routed, approvals are granted, data is governed, and exceptions are handled.
AI models reason. Enterprise platforms execute, remember, and remain accountable. As systems become more agentic, this distinction becomes more important, not less.
Agentic AI still needs scaffolding
One of the quieter ironies of the current moment is that many of the SaaS companies being punished by markets are best positioned to make agentic AI work in practice.
Take ServiceNow. Its value doesn’t come from knowing what to do, but from knowing how work actually moves across Information Technology (IT), Human Resources (HR), and customer operations, including who can trigger actions, what needs approval, and what must be logged. An AI agent operating inside ServiceNow has context, guardrails, and a clear path to execution.
Similarly, Salesforce doesn’t just store customer data. It governs how sales, service, and marketing actions are coordinated, measured, and audited. Agentic AI layered on top of that system doesn’t replace it; it becomes more effective because the system already exists.
Agentic systems don’t live in isolation. They need workflow context, permissions, integration maps, and failure-handling logic. That scaffolding is precisely what mature SaaS platforms already provide.
The enterprise constraint markets keep underestimating
There is another structural constraint that markets repeatedly overlook: compliance.
Large enterprises do not deploy systems that cannot be audited, governed, and controlled. Agentic AI introduces new risk vectors such as autonomous actions, probabilistic outcomes, and decision opacity. For F2000 companies, these are not theoretical issues. They are operational and regulatory realities.
Compliance, governance, and accountability cannot be vibe-coded. They have to be designed into platforms, tested over time, and trusted across business and risk teams. This reality strongly favors incumbents with embedded controls over thin, model-centric abstractions, no matter how compelling the demo.
We’ve seen this movie before in the IT BPS market
The current SaaS panic closely mirrors what happened in the services sector a few years ago. When generative AI first took off, IT services and Business Process Services (BPS) were widely assumed to be on the brink of collapse. Coding would be automated. Support work would disappear. Labor-based models would implode.
What followed was not extinction, but recalibration. Enterprises still needed services that were delivered differently. Providers with deep process knowledge adapted fastest, embedding AI into real operating models. Those without differentiation struggled. The sector evolved rather than vanished.
The SaaS market is going through a similar moment. AI is real. Disruption is real. But the pace and shape of change are being overstated.
From growth-at-all-costs to value-that-has-to-prove
Even before agentic AI entered the conversation, the direction of travel was clear. SaaS was moving away from growth-at-all-costs toward value that has to show up in production. Feature breadth mattered less. Utilization, outcomes, and pricing discipline mattered more.
Agentic AI doesn’t disrupt this shift. It accelerates it. As intelligence becomes cheaper and more abundant, the burden of proof shifts decisively to the platform. SaaS companies that cannot clearly demonstrate how AI improves outcomes inside real workflows will find it increasingly difficult to defend renewals, pricing, and expansion.
Where optimism is warranted
Agentic AI has the potential to make enterprise software more valuable, not less. It can reduce manual handoffs, accelerate decision-making, and elevate platforms from Systems of Record (SoR) to Systems of Action (SoA). For SaaS products embedded at critical decision points, the opportunity is real.
In many cases, AI will deepen platform stickiness rather than erode it.
Where caution is necessary
At the same time, this transition will not be kind to everyone. SaaS products that are lightly embedded, poorly differentiated, or sold primarily on promise rather than utilization will face sustained pressure. AI compresses feature advantage quickly and exposes shallow value propositions faster than before.
The bar is rising, and not all vendors will clear it.
What this means going forward
It’s worth separating valuation pain from operational reality. Much of the recent sell-off reflects a repricing away from 2021-era expectations rather than a sudden collapse in enterprise demand. Slower growth feels like stagnation only when markets remain anchored to an abnormal baseline.
For SaaS leaders, the implications are clear:
- Agentic capability needs to be anchored to workflows rather than features
- Governance and compliance need to be productized, not positioned as afterthoughts
- Value needs to be articulated in outcomes rather than access
The more important question isn’t whether AI agents will disrupt SaaS, they will. The question is how that disruption plays out. The answer is unlikely to resemble overnight disintermediation. It will look more like governed adoption, hybrid models with humans in the loop, and platforms evolving into control planes for agentic work.
Markets reacted as if intelligence alone replaces systems. Enterprises know better. Intelligence without governance is a liability. Autonomy without accountability is risk.
The Anthropic launch was a wake-up call and not a death knell. We’re not watching SaaS die. We’re watching it enter a more serious, more disciplined, and ultimately more durable phase of its evolution.
If you enjoyed this blog, check out, Enterprise AI adoption is rising, but the conviction is still missing – Everest Group Research Portal, which delves deeper into another topic relating to AI.
If you have any further questions and would like to discuss SaaS, Anthropic and the ever-changing world of AI, please contact Sharang Sharma ([email protected]) and Chirajeet Sengupta ([email protected]).