In technology circles, one of the most intriguing questions being debated today is whether AI agents will spell the end of traditional software as a service (SaaS). Leaders like Satya Nadella have argued that the rise of agentic AI—tools capable of automating and orchestrating tasks across systems—could upend the SaaS model entirely. Others believe that AI will simply enhance SaaS, evolving it but leaving its core value proposition intact.

Listening to this debate, I’ve found myself reflecting on what past technology disruptions can teach us about the answer.

One lesson history teaches us is that technological revolutions are rarely binary. Transitions don’t typically lead to total replacement. Instead, they create ecosystems marked by heterogeneity—a mix of old and new models, each finding its niche.

Take client/server computing, for example. It was supposed to eradicate mainframes. Yet today, many of my clients still rely on mainframes as a critical part of their infrastructure. Similarly, cloud computing was hailed as the replacement for on-premises systems, but hybrid models have become the norm. Businesses now balance cloud and on-premises architectures to optimize flexibility, control, compliance, security, and cost.

Perhaps my favorite example, given my personal history, is the humble PC. Mobile devices, tablets, and other form factors were heralded as the “death of the PC.” Yet here I am, typing this blog on a laptop. PCs remain the most essential devices for productivity-focused tasks, coexisting alongside mobile phones, tablets, and myriad other Internet-connected tools.

There are many more examples—programming languages that coexist rather than replace one another, network protocols that specialize in serving different layers, and even AI itself. AI has long been touted as a replacement for humans but has instead become an augmentative force. Terms like "human in the loop" and "human on the loop" are now mainstream in business conversations.

To be clear, these examples don’t guarantee resilience for incumbents. The future is inherently uncertain, and this debate highlights some of the most fundamental questions about how technology evolves, what customers value, and the essence of strategy in the technology industry.

Still, these examples do reveal a pattern: Transitions lead not to extinction but to transformation, adaptation, and coexistence. Will AI agents and SaaS follow a similar trajectory?

I would argue that it depends.

AI agents are undeniably disruptive. Their ability to automate workflows, connect disparate systems, and deliver outcomes without user intervention is a direct challenge to SaaS vendors. If AI can replace the “business logic” layer that SaaS applications provide, then the SaaS layer survives only if it creates unique and indispensable value for customers. SaaS vendors must innovate, focusing on what AI agents alone cannot offer—industry-specific expertise, deeper integrations, or bespoke capabilities tailored to complex workflows. This is the essence of software strategy.

For horizontal AI providers—the builders of these agentic tools—the challenge is equally steep. AI agents must absorb the nuanced use cases that SaaS platforms have spent years refining. To succeed, they need to establish trust and transparency with customers. Businesses won’t adopt tools they can’t control or verify, and the challenge of adapting AI to highly specific workflows across industries and inside companies is nontrivial. Historically, no one-size-fits-all solution has met the demands of a diverse, global economy.

But what if both sides rise to the occasion? SaaS vendors might embed AI deeply into their platforms, offering hybrid solutions that leverage the intelligence of AI agents while retaining the specialized strengths of SaaS. That is exactly what many technical roadmaps suggest they will attempt—though it will require a willingness to disrupt their own businesses from within.

Meanwhile, agentic providers could enhance their adaptability and transparency, creating tools that feel both trustworthy and customizable. In doing so, they could dismantle software layers that don’t create real customer value or lack competitive advantages (e.g., network effects, data moats, regulatory readiness).

If all of this happens, we won’t see collapse but convergence—a heterogeneous landscape where AI and SaaS complement and compete in various ways.

The lesson from history is clear: Transitions expand ecosystems rather than replace them outright. But that doesn’t mean the future is certain. The disruptive potential of AI agents places a heavy responsibility on SaaS vendors to prove their relevance in an AI-first world. Likewise, the adoption of agentic systems depends on their ability to handle real-world complexity without sacrificing trust or control.

The next chapter in this story will be how businesses, industries, and users respond to these shifts. It’s not a simple choice between AI and SaaS—it’s about the creative ways they adapt and coexist in a future that, as always, will be richer and more complex than we can predict.