This prediction is dominating conversations in the tech industry; AI will kill off SaaS. It’s no longer a question, but said with certainty.
Oversimplification rarely comes with strong evidence to support it though. In this instance, it demonstrates a misunderstanding of what SaaS really is. The real question is rather, which parts of SaaS is AI actually replacing (or likely to replace) in the coming years?
So, let’s take a step back. What is and what is not SaaS?
Obviously, SaaS has so many dimensions that answering this question is not straightforward. So, let’s take it the other way around. What is not SaaS?
SaaS is not one single thing, like a monolithic block that could be easily replicated by a simple act of will. It’s a combination of user experience, business logic, domain expertise, data, integrations, governance, security, and execution. Academic research in digital innovation reaches a remarkably similar conclusion. The New Organising Logic of Digital Innovation argues that digital products are built upon a layered architecture composed of loosely coupled layers rather than a single monolithic entity, creating different sources of value and innovation across the stack.
My belief is that AI cannot affect all of these equally. So, which parts of SaaS are actually at risk because of AI?
The first thing that comes to mind is the friction embedded in many SaaS experiences. Many of us have experienced the countless clicks, endless menus, the overuse of drag-and-drop form builders, overly complex searches and dashboards from multiple perspectives, and repetitive workflows… and that’s often just the start.
All of this is highly exposed to AI. The new technology gives users the ability to skip these steps and express the outcome they want instead of navigating the SaaS application. In other words, LLMs allow users to formulate the outcome directly rather than learning software navigation models that usually vary from one application to another. It is obviously much easier to ask your favourite AI companion to review all contracts expiring by a given date and perform a holistic analysis of churn risks.
From this perspective, AI is astonishing, and it represents a real shift. But ending the analysis here is quite premature and ultimately reflects a misunderstanding of what SaaS really is.
Removing friction doesn’t eliminate the system behind it. In fact, it’s the complete opposite. The economic literature on information systems has long suggested that most business value does not reside in the software interface itself but in the organisational capabilities, processes, and complementary assets surrounding it.
Often, the more AI you add, the more the value of the underlying data and system behind it increases. The more capable AI becomes, the more it depends on trusted enterprise platforms to access reliable data, enforce permissions, execute transactions, maintain compliance and orchestrate increasingly complex business processes. Those capabilities don’t emerge by accident. They are built over years through systems of record, business logic, domain expertise, governance, security and trust.
That said, a more nuanced statement may be that AI does not replace all of SaaS equally. Ironically, the companies most threatened by AI may not be traditional enterprise SaaS vendors. I would argue that they are the companies whose primary value proposition was simplifying tasks, on a SaaS-like approach, that AI can now perform natively. For example: writing, summarising, searching, classifying, documenting or generating content. These capabilities are rapidly becoming commodities or, at the very least, they cannot sustain value on their own if there is little else behind the scenes.
This distinction is important because the exposure is often task-specific rather than company-specific. It reflects recent work on the idea that AI assistance can dramatically improves performance on some knowledge tasks while remaining unreliable on others. A process described by Havard academics as the ‘jagged technology frontier’.
However, it is important to understand that the value of a SaaS has never been just the interface or the dashboards. It is what lies behind. It is what users don’t see:
i) the underlying data model(s),
ii) the workflows refined over years,
iii) the integrations connecting dozens of business systems,
iv) the security model,
v) the compliance frameworks,
vi) the operational resilience,
vii) the domain expertise built alongside users facing real business pain points,
viii) the transparency and explainability of the outputs.
In the same kind of thinking, recent research on digital platforms suggests that Generative AI transforms the interaction layer of SaaS platforms much more than the underlying platform itself. They can handle context, unstructured requests and generate novel outputs, which ‘create[s] new forms of value mediation between platform participants’.
And this is precisely why the ‘end of SaaS’ narrative falls short. These assets become even more valuable when AI agents begin acting autonomously on behalf of users.
It might sound like a paradox but the more autonomous AI becomes, the more critical trusted enterprise foundations become. Some recent research mentions it this way: “Security and trustworthiness become essential prerequisites for the safe deployment of agentic AI.”
Conversely, this is exactly where AI agents become agents of chaos when they operate without these foundations. AI agents may amplify errors at scale rather than eliminate them. This risk is increasingly documented in emerging research on autonomous agents.
That’s why I don’t believe at all that AI and SaaS are necessarily on a collision course. Rather, I believe they are converging.
Hence, the more precise question is not whether AI will kill SaaS, but which layers of SaaS will remain differentiated once AI becomes embedded in SaaS at a large scale. The answer may be as simple as this: interfaces, navigation, reporting and content generation may become commodities quite fast. On the flip side, trusted data, workflows, governance, integrations, compliance, security and execution capabilities may become even more strategic.
The direct consequence is that SaaS products whose value proposition is limited to their interface or dashboards may disappear with it. But the platform will remain.
I would even argue that AI could become the greatest growth engine the SaaS industry has ever seen. Not because it replaces SaaS, but because it finally unlocks the value that has always been hidden inside it. That being the decades of business logic and enterprise knowledge.
In other words, AI doesn’t create business logic. It consumes business logic.
The post The Prediction that AI will Kill Off SaaS is Premature. A Better Question is Which Elements are Converging? appeared first on The Fintech Times.
