For the better part of two decades, software followed a familiar script. Venture capital flowed into SaaS companies, founders raced to define new categories, and winners built enduring franchises by owning those categories outright.
If you could name the problem and frame the solution convincingly enough, you could often secure market leadership before the product had even fully matured.
That playbook is now showing its age.
Today, venture capital – particularly in the United States – is increasingly concentrated in artificial intelligence. Portfolios that once brimmed with SaaS bets are being reweighted toward AI infrastructure, foundation models, and adjacent tooling.
The shift is not subtle.
It reflects a deeper change in how value is created, captured, and defended in software.
SaaS, for all its success, has matured. The category-defining era of the 2010s produced giants across nearly every business function – CRM, collaboration, commerce, finance, and beyond. Companies like Salesforce, Slack, and Shopify didn’t just build products, they built categories.
But those categories are now crowded, well-understood, and difficult to disrupt in meaningful ways.
Most new SaaS startups today are not inventing entirely new markets. They are iterating – building incremental improvements, targeting niches, or offering slightly better user experiences within established categories.
While these businesses can still succeed, they rarely command the same excitement or valuations as their predecessors.
Artificial intelligence changes the equation.
Unlike SaaS, which is typically verticalized and category-specific, AI is inherently horizontal. When a major AI provider releases a new model or capability, the effects ripple across dozens of industries simultaneously. A single advancement can improve customer support tools, coding platforms, marketing systems, and knowledge management software all at once.
This has profound implications. It means that innovation is no longer neatly contained within category boundaries. Instead, it flows across them, often eroding the differentiation that SaaS companies rely on.
For startups, this creates a structural challenge. Many are now building on top of foundational AI platforms – large models developed by a small number of dominant players. While this enables rapid development, it also introduces dependency. If the underlying platform evolves or decides to integrate a feature directly, the startup’s advantage can disappear overnight.
History offers a parallel.
In the early days of personal computing, software developers built applications on top of operating systems controlled by a handful of companies. Those platform owners were initially happy to encourage third-party innovation – it drove adoption of their core products. But over time, as growth slowed, they began to absorb the most valuable features into their own ecosystems, squeezing out independent developers.
A similar dynamic is emerging in AI.
This shift is also reshaping how investors evaluate companies. During the SaaS boom, the central question was straightforward: can this company dominate its category?
Today, that question feels insufficient. Instead, investors are asking what proprietary advantage a company truly possesses. Is it unique data? Exclusive distribution? A defensible model architecture?
Without one of these, differentiation is fragile.
The nature of competitive moats is changing as well. SaaS companies historically relied on workflow lock-in, high switching costs, and deep integrations. Sales teams and ecosystems reinforced their position. AI companies, by contrast, compete on different dimensions: access to high-quality training data, computational resources, and the ability to maintain context and performance at scale.
Speed is another critical difference.
SaaS categories often took years to form and mature, allowing companies to grow steadily and investors to compound returns over time. AI categories can emerge and saturate in a matter of months. What begins as a breakthrough quickly becomes crowded, with multiple players offering similar capabilities.
This rapid cycle leads to what might be called “category inflation.” Nearly every startup now positions itself as an AI platform, regardless of how much underlying technology it actually owns. In many cases, these offerings are thin layers built on top of existing models – useful, but not deeply defensible.
This phenomenon is particularly visible in ecosystems where expectations around acquisitions are high. Some founders assume that building a clever application on top of a major AI platform will naturally lead to lucrative buyouts. While such outcomes do occur, they are exceptions rather than the rule, and often depend on networks, timing, and strategic fit rather than just technical novelty.
Meanwhile, capital is concentrating at the extremes. Foundational AI companies and infrastructure providers are attracting enormous investment, while application-layer startups – especially those without clear differentiation – are finding funding harder to secure.
This marks a reversal from the SaaS era, when application companies captured most of the value and infrastructure was relatively commoditized.
Hence, the “software switcheroo.”
So where do new opportunities lie?
One emerging answer is AI-native workflows. Instead of embedding AI into existing tools, a new generation of companies is rethinking workflows from the ground up. These systems are increasingly autonomous. They can write code, conduct research, analyse data, and even make operational decisions.
This represents a fundamental shift – from software as a tool to software as a collaborator.
But with that shift comes new challenges. Trust becomes central. Organizations are no longer just adopting software; they are delegating responsibility. Questions of governance, oversight, and accountability move to the forefront.
As a result, the next major battleground may not be the models themselves, but the operational layer that manages how AI is deployed across organizations. The companies that define this layer – how AI systems are developed, monitored, and integrated into business processes – could hold significant power.
This also creates an unexpected opportunity for large consulting firms. For decades, they have helped organizations implement and optimize software systems. Now, as businesses grapple with how to integrate AI into their organizations, that expertise becomes newly relevant. However, success will depend on how quickly they can adapt their consulting capabilities to this new paradigm.
The broader lesson is clear
The future of software will not be shaped solely by who builds the most advanced models. It will be defined by who determines how those models are used – how they fit into workflows, how they are governed, and how they create added value in organizations.
The SaaS era was about owning categories. The AI era may be about orchestrating intelligence.
And that is a very different game.