Vijay Raina has spent years at the intersection of enterprise architecture and software strategy, observing how digital products evolve from simple tools into complex ecosystems. As the Chief Executive Officer of Sixthfin since early 2026, he has witnessed firsthand the shift from monolithic software designs to the layered architectures that define modern SaaS. His expertise lies in understanding that software is far more than just the buttons we click; it is a sophisticated stack of domain expertise, data integrations, and governance frameworks. In an era where many are quick to predict the total obsolescence of software-as-a-service at the hands of artificial intelligence, Raina offers a more grounded perspective. He argues that while AI will certainly transform how we interact with technology, the “invisible” layers of the SaaS stack are becoming more critical than ever. This conversation explores the convergence of AI and SaaS, the risks of autonomous agents, and why the most valuable parts of a platform are the ones users never see.
The conversation around AI often suggests it will completely replace traditional software, but you’ve argued that this perspective stems from a misunderstanding of what SaaS actually is. How do you distinguish between the visible user interface and the underlying layers like business logic and governance?
To understand where we are in 2026, we have to stop viewing SaaS as a single, monolithic block that can be easily replicated. In reality, modern digital innovation is built upon a layered architecture where different levels of the stack provide distinct sources of value. When people say AI will “kill” SaaS, they are usually only looking at the interaction layer— the menus, the endless clicks, and the drag-and-drop form builders that have defined our workflows for a decade. These elements represent friction, and AI is exceptionally good at removing that friction by allowing users to express an outcome directly. For instance, instead of navigating a complex dashboard, a user can simply ask an AI companion to perform a holistic analysis of churn risks for all contracts expiring by a specific date. However, the software doesn’t end there; beneath that interface lies the business logic, the refined workflows, and the compliance frameworks that ensure the output is actually valid. The interface is just the tip of the iceberg, and while AI might melt the tip, the massive structure of data models and security protocols beneath the surface remains the essential foundation.
If the interface and reporting layers are becoming commodities because of AI’s ability to generate content and formulate outcomes, which specific types of SaaS companies are most at risk in this new landscape?
The companies facing the greatest threat are those whose primary value proposition was simply simplifying tasks that AI can now perform natively. Think about the tools we’ve used for years to write, summarize, classify, or document content—the SaaS-like applications that focused on these specific knowledge tasks. These capabilities are rapidly becoming commodities because an LLM can handle them without needing a specialized third-party interface. Research into the “jagged technology frontier” shows that while AI assistance dramatically improves performance on these types of tasks, it remains unreliable on others that require deep context or organizational memory. If a company’s only “moat” was a sleek dashboard for a task that AI now handles with a single prompt, that company is in trouble. The value has shifted away from the “task” itself and toward the platform that can orchestrate that task within a trusted enterprise environment, complete with the necessary permissions and operational resilience.
You’ve mentioned the “paradox” that the more autonomous AI becomes, the more critical traditional SaaS foundations like security and governance become. Why do AI agents risk becoming “agents of chaos” without these structures?
It is a significant concern because as AI agents begin to act autonomously on behalf of users, they have the potential to amplify errors at scale rather than eliminate them. Without a robust system of record and a clear governance model, an autonomous agent is essentially operating in a vacuum without any guardrails. This is why trusted enterprise foundations are more strategic now than they were even a few years ago; security and trustworthiness are the absolute prerequisites for safe deployment. If an agent tries to execute a transaction or move data between dozens of business systems without following the established security model or compliance frameworks, the results can be catastrophic. We aren’t just talking about a typo in a summary; we are talking about systemic failures in business logic. Those capabilities—the ability to enforce permissions and maintain transparency—don’t emerge by accident or by training a model; they are built over years of domain expertise and engineering.
As we look at the internal mechanics of a platform, what are the “invisible” assets that you believe will remain differentiated and even gain value as AI becomes more deeply embedded?
The real value of SaaS has always been what lies behind the curtain, and those assets are only becoming more valuable as AI begins to “consume” them. First and foremost are the underlying data models and the integrations that connect various business systems into a cohesive whole. Then you have the workflows that have been refined over years of addressing real-world business pain points, along with the operational resilience that ensures a system stays up and running under pressure. There is also the matter of explainability; in a professional setting, you cannot just have an output—you need to know why that output was generated and if it meets specific compliance frameworks. These elements—the domain expertise, the security models, and the integrations—are not things AI creates; they are things AI depends on. In many ways, AI is the engine, but the decades of business logic and enterprise knowledge stored within the SaaS platform are the fuel that makes the engine useful.
If we accept that AI and SaaS are converging rather than colliding, how should enterprise leaders rethink their software investments over the next couple of years?
Leaders need to move past the binary choice of “AI vs. SaaS” and start asking which layers of their software stack are truly strategic and which are just overhead. We are moving toward a reality where the navigation and reporting functions of software are essentially a given, so the focus must shift to the execution capabilities and the quality of the data. I would argue that AI could be the greatest growth engine this industry has ever seen because it finally unlocks the value that has been hidden inside these platforms for years. Instead of being buried under a mountain of clicks, that business logic can now be accessed and utilized through a conversational layer. The direct consequence is that products limited to being “just a dashboard” will likely disappear, while the platforms that own the data and the workflows will become the central nervous system of the enterprise.
What is your forecast for the SaaS industry as it becomes fully integrated with agentic AI systems?
My forecast is that we are entering a period of radical simplification at the surface level, coupled with intense sophistication at the foundation level. By the end of 2026 and into 2027, we will see the “interaction layer” of almost every major SaaS platform completely transformed to handle unstructured, natural language requests. This will create new forms of value mediation between different participants in a platform, making the software feel less like a tool and more like a collaborator. However, this shift will also trigger a massive “flight to quality” where enterprises consolidate their spending around a few trusted platforms that have the most robust security and the deepest integrations. We will see the “end of SaaS” as a collection of fragmented tools, but the rise of the “Enterprise Operating System” where AI-driven agents operate seamlessly within the safe, governed boundaries of the underlying platform. The winners won’t be the ones with the best AI models, but the ones who provide the most reliable environment for those models to actually get work done.
