Why Markets Are Punishing SaaS and What the Path to Survival Looks Like

In recent months, a troubling trend has been resonating across social media and in industry publications: a decline in the stock prices of technology leaders providing Software as a Service (SaaS). Investors who once admired the scalability of SaaS are becoming skeptical. Why is this happening precisely at the peak of the artificial intelligence boom?

At its core, we are confronting two main theses that are shaking the foundations of the existing business model.

Thesis 1: The Democratization of Code, or “I’ll Just Prompt It Myself”

The first argument is based on the idea that AI, through the democratization of development, now enables code to be generated so quickly and cheaply that the value of finished software is declining. If, thanks to AI, virtually anyone can “code,” companies will supposedly prefer to build their own tailored solutions rather than pay for expensive seat-based licenses.

The reality, however, is that code represents only the tip of the iceberg. Anyone who has ever built a robust enterprise system knows that software is primarily about performance, security, deep integrations, reporting, audit trails, and regulatory compliance.

No one is going to “prompt” SAP into existence, because behind it stand decades of domain expertise and complex architectural decisions. Someone might manage to generate a simple Request Fulfillment tool, but do they truly want to maintain and integrate it over the long term? Does it really make sense for every company to build its own unique code for generic processes? It is more likely that the nature of SaaS tools will evolve. The traditional user interface (UI) will lose importance, and the key differentiator will become a high-quality interface for AI agents, enabling the autonomous engagement of “digital employees.”

Thesis 2: The Marginal Cost Trap

The second thesis strikes at the economic heart of SaaS. Investors loved software because the marginal cost of each additional user was virtually zero after development, creating the famous profit “hockey stick.”

The rapid adoption of AI assistants has disrupted this model. Due to token pricing and computational costs, AI embedded in software behaves more like expensive infrastructure. Unless computation costs decrease dramatically, the economics of scaling cease to function as before. One potential response is the use of local models, but these require massive investments in proprietary infrastructure and tooling. In an environment where the AI landscape changes every three months, it is difficult to justify such investments for mere “assistants” whose real value users are increasingly questioning.

The Way Forward: From Selling Licenses to Selling Outcomes

How do we escape this trap? The solution lies in changing the economic model. We must stop selling AI as software (a tool) and start selling it as completed work.

This model resembles outsourcing: value does not scale with the number of licenses sold or hours billed, but with the delivered outcome and the assumption of responsibility for it. While today’s SaaS typically sells tools for ten accountants, in the AI era we will sell a guarantee of fully processed payroll.

If we can build a reliable digital service, we can sell it as a comprehensive result. This model scales far better than merely renting software capacity—because the efficiency gains delivered by AI can be monetized as direct value for the customer.

Confirmed by Experts

Analysts at McKinsey & Company see the situation similarly. In their recent report, “Upgrading software business models to thrive in the AI era,” they confirm that the era of per-seat software sales is coming to an end. The winners will be those who redefine their business models and move from selling tools to delivering measurable efficiency and tangible business outcomes.