Is the golden era of SaaS coming to an end? Why and how AI might (or might not) take over

Voices from the startup and investment scenes suggest that this business model, where products are sold on a subscription basis, might be nearing its demise. The rise of AI is expected to significantly weaken the competitive defenses of software players—along with their profit margins. What could this mean?

There’s talk that “moats” (a term popularized by legendary investor Warren Buffett to describe competitive advantages) have never been shallower. As a result, investors and founders must question whether SaaS remains the best option for their resources. However, as with most popular tech memes, the reality is more complex.

AI as an Agent of Change

Whether you ask OpenAI’s Sam Altman, Anthropic’s Dario Amodei, or the top AI investors, the answer is consistent: the future is in AI agents. Imagine these agents as synthetic versions of employees — capable of planning and executing tasks independently or under human supervision. They come equipped with memory, reasoning skills, and access to systems and databases.

AI agents will be able to read your device screens, navigate browsers, and communicate with other agents. For instance, Replit’s development agent can create an app based on a simple verbal request, delivering the first version in no time. Similarly, Fin, an AI agent for customer support, goes beyond handling basic queries; it efficiently routes complex issues to human agents when necessary.

Among the major tech players, Microsoft has embraced AI agents most aggressively. Google, like Anthropic recently, plans to introduce artificial intelligence capable of operating computers as humans do. According to recent reports from Bloomberg, OpenAI is set to launch its agent tool in January next year.

More work performed by AI means a reduced need for people, but also less software designed as tools for those people. After all, products are merely software representations of workflows. So what happens when AI takes on a substantial share of tasks and processes—or, more importantly, radically transforms them with its capabilities?

Products that support human workflows may lose their purpose or much of their significance. And if there’s one thing these products have in common, it’s the subscription-based software-as-a-service (SaaS) model. Some might argue that existing SaaS players rapidly incorporate AI features to fortify their market positions. While this is true to some extent, it’s not the whole story.

Adding a few AI features is built around the concept of a copilot—an AI assistant that helps people with their work. Incorporating functions like summarization or text generation into existing products is neither as difficult nor risky as pursuing full automation and cannibalizing one’s business model. A model that typically generates gross margins exceeding 70%, is sold as a subscription, and is priced per seat. A model that, over the past decade, has created numerous unicorns and become the most common target for founders and venture capital investors.

For providers, SaaS is perhaps the best they could wish for, thanks to its high margins. However, for customers and their budgets, the situation is quite different. The trend of reducing software subscriptions and minimizing SaaS spending may have been triggered by COVID, rising interest rates, and economic uncertainty. However, the real challenge for the future of SaaS lies in artificial intelligence.

The competitive advantage of software players has largely been based on the complexity and cost of development. However, development costs will drop dramatically if AI becomes the primary driver of software creation in the coming years. The ease of entry for new market competitors—as well as internal competitors created by customers themselves—points toward the commodification of software and the prospect of lower or even non-existent margins.

A prime example is Klarna, which publicly moved away from costly tools like Workday and Salesforce. The fintech giant claims that, with the help of AI, it can build the necessary tools internally and achieve significant cost savings. While this move can also be viewed through the lens of its planned IPO, it has made a strong impact. A notable shift in the thinking of many IT leaders and technology strategists is how they calculate whether to buy solutions or develop them internally. This shift suggests that building solutions internally may now make more sense than purchasing them—or continuously paying for subscriptions.

A few reasons why it’s more complicated:

First: “Software doesn’t make products. Products don’t make businesses.” This comment from Rasty Turek, founder of the startup Pex, under the online essay The End of Software, clearly explains why artificial intelligence's impact as a software disruptor is frequently exaggerated. The competitive defenses of software go beyond the ability to generate code; truly effective moats are built around the following:

1. Network effects: Your service improves as more people use it or as it works with larger datasets.
2. Economies of scale: You can deliver the same or greater value at a lower cost because you’re large and it costs you less.
3. Intellectual property: For example, proprietary AI models trained on a unique dataset.

Other forms include advantages in distribution, a wide range of features, integrations, platformization, and even the difficulty of switching providers (vendor lock-in). One of the most commonly repeated arguments against SaaS has a notable weakness. It argues that the competitive edge provided by development complexity has disappeared. However, that was never a solid defense to begin with.

Second: The potential of in-house development is overestimated. While Klarna might be an inspiring example of cost reduction, we shouldn’t expect internal teams to start creating solutions on a large scale that replace external software providers. SaaS subscriptions guarantee customers ongoing support, continuous development, and updates. Building a simple app with a dashboard is one thing, but companies often need secure, integrated, and robust solutions that require maintenance.

Third: Reliable AI agents—even those capable of generating software at a human level—are not here yet. The CEO of Anthropic says it will take two more leaps in model scale, which could mean GPT-6 or its equivalent. The lower reliability of AI solutions makes companies cautious, leaving traditional SaaS for existing workflows as the best option for now. The clock is undoubtedly ticking for current players, but they still have time to follow Intercom’s lead and position themselves as the ones who guide their customers into the future.

Fourth: Any change, even one that promises cost savings and increased revenue, involves significant effort for customers. Implementing automated solutions inevitably demands transformation not only of the technology stack but also of the business model, organizational structure, culture, and workforce. Analyst Benedict Evans highlights, through the history of automation, that even seemingly straightforward transitions to the cloud have significantly lagged behind expectations. Introducing AI agents into corporate processes could take much longer—and not because of their insufficient capabilities.

In 2011, Marc Andreessen wrote that software is eating the world. As we near the end of 2024, this remains true. Declaring the death of the SaaS model is premature and shortsighted. Existing players still have ample time to transform, and new startups are already building for an AI-driven world. We can expect pricing models to shift, focusing on usage or outcomes rather than simply charging per user.

New interfaces and tools will emerge, enabling seamless collaboration between humans and artificial intelligence. AI will transform new industries and unlock a wide range of new use cases, leading to an exciting evolution of software and the AI-driven SaaS model.

This article was written for Czechcrunch.cz (in Czech).