Vijay Raina, a leading voice in enterprise SaaS technology and software architecture, has navigated the shifting tides of the software industry for decades, from the early days of monolithic databases to the current explosion of agentic programming. As “vibe-coding” and natural language prompts empower individuals to build their own bespoke tools, Raina offers a critical perspective on why the death of traditional Software as a Service is greatly exaggerated. He emphasizes that while AI can help a single user create a custom skill in seconds, the true challenge of the modern enterprise lies in collaboration, data integrity, and the transition from human-centric dashboards to machine-ready APIs. In this discussion, we explore the friction between personal productivity and organizational cohesion in the age of AI agents.
When individuals build custom AI tools like personal CRMs using natural language, data schemas often diverge. How do we reconcile the immediate efficiency of a salesperson building their own tool with the long-term organizational nightmare of fragmented, incompatible data?
Imagine walking through a busy sales office where every person at every desk has become their own independent software developer. One salesperson uses a prompt to build a CRM that tracks addresses in a single text field, while the person sitting right next to them creates a version that breaks locations down into city, state, and zip code. They might be using different backends entirely—one on a local SQLite instance and another on a corporate Oracle database—and suddenly, the seamless flow of information that a company needs to survive is gone. This is the “silo problem” on steroids; while each individual feels 10% or 20% more productive in their specific workflow, the organization loses the ability to generate a unified report or track overall success. We have to realize that work is a team sport, and if everyone is playing by their own digital rules, you don’t have a company—you have a collection of isolated islands that can’t talk to each other.
You have mentioned that SaaS companies are excellent at building dashboards for humans, but agents require raw, structured data. How does this shift the fundamental design of enterprise software from a visual experience to a technical one?
For the last twenty years, the “gold standard” of SaaS has been the dashboard, a beautifully compressed visual representation of data that allows a human manager to glance at a screen and understand their pipeline. But an AI agent doesn’t need a pretty chart; in fact, the dashboard is a barrier because it hides the raw, “structured state” that an agent needs to actually perform tasks. We are moving toward a reality where the “Salesforce of Agents” won’t look like a website at all, but rather a robust set of APIs that provide relationship graphs, permissioned memory, and machine-readable sales playbooks. SaaS companies will thrive not by how they present data to eyes, but by how they deliver high-fidelity, accurate data to the machines that will take care of the “compression” themselves. It is a fundamental pivot where the primary customer is no longer a person clicking buttons, but an agent executing a sequence of intents based on reliable, structured information.
There is a significant gap between an individual “vibe-coding” a tool and a team using it collaboratively. What specific infrastructure is currently missing to make these agentic tools truly useful in a corporate environment?
The current state of agentic programming is incredibly solitary, and we are missing the basic “plumbing” required for team-wide adoption, such as robust sharing, versioning, and testing mechanisms. Right now, telling a non-technical user to “just put a markdown file in a corporate GitHub” is a recipe for high friction and total abandonment because the average professional doesn’t want to manage Git branches or merge conflicts. We need a private marketplace or an administrative dashboard that wraps these complex developer workflows into a simple, consumer-grade experience. Furthermore, we haven’t even begun to master “evals” for these skills; if a tool is going to write a financial report or a sales projection, the company needs to know it won’t drift or backfire over time. Without these guardrails—security against prompt injection and clear version control—these personal AI tools will remain interesting experiments rather than the backbone of a department.
If the democratization of programming allows anyone to create a custom tool for a few dollars in tokens, what is the ongoing value proposition of paying for a massive, feature-heavy subscription like Salesforce?
It is easy to look at a giant platform and say, “I only use 5% of these features, so I’ll just build my own version,” but that ignores the “hidden value” of the bundle. That other 95% of features often contains the one specific thing you don’t realize you need until a crisis hits, or a unique way to extract value from data that you haven’t even thought of yet. SaaS companies offer a standard of record and a shared language that individual tools simply cannot replicate; they provide the “connective tissue” that allows different departments to share metrics and goals. There is a deep, inherent value in buying a pre-built ecosystem that has already solved the problems of security, interoperability, and scalability. Even if we had the tools to code everything ourselves 30 years ago, we still would have gravitated toward shared platforms because the friction of maintaining 500 different “personal” tools is far more expensive than a monthly subscription fee.
Large, established tech companies are often criticized for moving slowly. In the face of rapid AI advancement, what is the biggest risk for these giants if they fail to adapt their “inertia” into a new direction?
In the world of physics, inertia and momentum are essentially the same thing, and for a company like Google or Salesforce, their massive size is both their greatest strength and their most dangerous weakness. They have the data and they understand the requirements of the enterprise better than anyone, but the “large institution” mindset makes it incredibly difficult to pivot as fast as the technology is moving. The risk isn’t that AI will replace them, but that they will be blindsided by smaller players who build “agent-first” infrastructures from the ground up while the giants are still trying to polish their legacy dashboards. If these companies don’t realize that their real future lies in providing high-quality APIs for agents rather than just UIs for humans, they will become the digital equivalent of a “system of record” that nobody knows how to talk to. They must recognize that the programming language of the future is English, but the underlying data architecture is still as complex as C++, and they need to lead that transition before someone else defines the new standard.
What is your forecast for the future of SaaS?
I expect that over the next five years, we will see a radical “unbundling” of the user interface where the API becomes the primary product, and the traditional SaaS dashboard becomes a secondary tool for occasional human oversight. We will move into an era of “Hybrid SaaS,” where the core platform acts as a high-integrity data hub that feeds hundreds of specialized, agentic skills tailored to individual user needs, yet all governed by a central set of rules. The companies that survive will be those that solve the “sharing and collaboration” problem for AI agents, creating a marketplace where custom-built skills can be safely distributed, versioned, and audited across an entire organization. Ultimately, SaaS isn’t going away; it’s just going “under the hood” to become the essential engine that powers a world of autonomous, intelligent agents.
