Vijay Raina is a seasoned authority on the evolving architecture of enterprise software, helping organizations navigate the transition from rigid application silos to fluid, data-centric ecosystems. As the boundary between software and data continues to blur, his insights into how artificial intelligence is stripping away the traditional advantages of legacy SaaS providers have become essential for modern tech leaders. In this conversation, we explore the strategic necessity of owning the user experience, the rise of agentic applications that function without traditional interfaces, and why the “semantics” of data—how it is understood and connected—has become more valuable than the code itself.
If data and semantics are becoming more valuable than the software itself as AI simplifies application development, how should enterprises restructure their tech stacks? What specific steps should leaders take to prioritize data context while maintaining a seamless user experience?
Enterprise leaders must pivot away from the “software-first” mentality that has dominated the industry for decades. Since artificial intelligence has made generating code and applications almost trivial, the real competitive weight now lies in the semantics—the specific meaning, definitions, and relationships within your data. Over the last two-plus years, the conviction has grown that owning the user experience is critical because that is where the context is captured and maintained. To prioritize this, leaders should implement a unified data layer that serves as the foundation for all applications, ensuring that when AI models interact with business logic, they aren’t hallucinating on fragmented or poorly labeled information. This transition requires moving beyond traditional warehousing and adopting tools that allow for deep, contextual understanding of every data point across the organization.
As companies move toward agentic applications that may lack traditional user interfaces, how do foundational technologies like Hybrid Tables and Postgres support these operational shifts? What metrics should developers track to ensure these new applications effectively manage both analytical and operational tasks?
The shift toward agentic applications is a fascinating evolution because these tools often operate autonomously to fulfill complex goals, sometimes without a visual interface for a human to click through. Foundational technologies like Hybrid Tables and Postgres are essential here because they bridge the gap between heavy, long-term analytical processing and the lightning-fast operational tasks required for real-time decision-making. Developers need to track performance metrics that measure how well these systems handle “bidirectional” flows, ensuring that an agent can read historical trends and write new operational data simultaneously without a bottleneck. It is no longer enough to have a fast database; you need a system that provides a consistent sensory experience for the AI agent, allowing it to act on the most current data available. By focusing on these hybrid capabilities, companies can support a new generation of apps that are as capable of “doing” as they are of “thinking.”
Given the risk of data fragmentation, how can organizations successfully implement a centralized data layer using zero-copy partnerships? What are the practical trade-offs when deciding whether to integrate with existing software ecosystems versus building custom, data-centric applications?
The goal is to stop the wasteful cycle of copying the same data into multiple different applications just because those apps want to “own” their own database. By utilizing bidirectional, zero-copy partnerships, organizations can allow their data to remain in a centralized layer while still giving various software providers the access they need to perform specific functions. The practical trade-off usually comes down to speed of deployment versus long-term control; integrating with an existing ecosystem is often faster, but building custom applications using platforms like CoCo or CoWork allows you to retain total ownership of the semantics. Most enterprises will find a middle ground, but the priority must always be avoiding the “silo effect” where data becomes trapped and unusable for broader AI initiatives. Successfully navigating this requires a disciplined approach to choosing partners who respect the integrity of your centralized data layer.
With the rise of AI models that allow software to be created more easily, how does a company’s unique data footprint serve as a competitive moat? Could you share an anecdote or scenario where owning the end-to-end user interaction with that data changed a business outcome?
Your company’s unique data footprint is the only thing a competitor cannot replicate with a clever prompt and a standard large language model. When a business owns the end-to-end user interaction, they capture nuances—the “why” behind a customer’s specific choice or a subtle supply chain delay—that simply aren’t present in generic, public datasets. I’ve seen scenarios where a firm used its deep data context to automate high-level workflows that previously required dozens of manual steps; because they owned the semantics, the AI could solve the problem with surgical precision rather than generic guesswork. This ownership creates a moat where the software feels like a natural, living extension of the business rather than a generic tool that anyone could buy off the shelf. We have seen this play out over the last two-plus years, where the winners are those who didn’t just buy a “bot,” but built an intelligence layer on top of their own proprietary history.
What is your forecast for the future of the SaaS industry as the focus shifts from software functionality to data ownership and AI-driven insights?
I expect the SaaS landscape to move toward a “headless” model where the primary value isn’t found in the buttons or the dashboard, but in the intelligence the platform provides over your specific data. We will likely see a significant thinning of the traditional application layer as centralized data platforms become the true operating systems of the modern enterprise. Successful companies will stop purchasing software based on its standalone features and start selecting partners based on how seamlessly they integrate into a zero-copy architecture. Ultimately, the industry will be divided between those who provide the “brain”—the data and semantics—and those who provide the “hands,” or the AI agents that execute tasks. This shift will favor the data owners, as the ability to generate software on the fly will turn the applications themselves into a commodity, while the data driving them remains the ultimate prize.
