Vijay Raina is a seasoned veteran in the SaaS landscape, having spent decades architecting the digital frameworks that underpin modern enterprise operations. As we navigate the complexities of 2026, his insights into the “invisible” layer of technology—where the traditional user interface fades in favor of agentic AI—provide a crucial roadmap for businesses struggling to maintain margins and relevance in a post-UI world. Today, we explore why the most powerful brands in software are choosing to disappear from our screens to become the essential, hidden plumbing of the global economy. This conversation delves into the structural shift of software into infrastructure, the radical dismantling of per-seat pricing models in favor of “tokenomics,” and the critical “harness” that businesses must build to ensure their data remains an asset rather than a liability.
The traditional software interface is increasingly described as a secondary gateway, with top executives admitting they rarely log into core applications anymore. How does this shift from a button-and-menu command structure to an agentic, prompt-based interaction change the fundamental value proposition of a SaaS giant?
The shift we are witnessing is nothing short of a total deconstruction of the user experience as we have known it for thirty years. When you hear that the leadership at a major firm like Salesforce—from the CEO to the CMO—has shifted their primary entry point to Slack or an AI prompt, you are seeing the end of the “UI as the product.” For a generation, sales and marketing professionals grew up on the wallpaper of specific dashboards and navigation bars; those interfaces were the physical manifestation of the brand. Now, as Patrick Stokes rightly pointed out, the UI is effectively being demoted to just one of many possible windows. The true product has migrated deeper into the stack, becoming the platform where data, metadata, workflows, and security permissions are encoded.
This is a profound transition because it means the value is no longer in how easily a human can click a button, but in how effectively an agent can access an API. We are moving into a world where SaaS vendors are content to slip beneath the prompt, allowing their capabilities to be accessed through someone else’s interface, like Claude or ChatGPT. It is a gamble that the architecture—the “Data 360” layer and the semantic connections built over 27 years—is robust enough to remain indispensable even when the logo is nowhere to be seen. You can feel the weight of this change when you sit in a place like San Francisco’s Moscone district during a major tech event; the branding is everywhere on the street, but when you actually sit down to work in a demonstration of “Claude Force,” the Salesforce logo is completely absent from the screen. It is an addictive, highly efficient way to work, but it detaches the user’s emotional connection from the software brand and reattaches it to the intelligence of the model.
If the value of software is shifting from the “presentation” to the “plumbing,” what does that mean for the profit margins of companies that have historically enjoyed 80% gross margins by selling the same code over and over?
The economic reality for SaaS firms is becoming significantly more turbulent as they transition into this “infrastructure” role. We’ve already seen a staggering 60% decline in median SaaS multiples from their peaks in 2021, and the agentic transformation is only going to accelerate that pressure. Historically, these companies built a product once and sold it on a per-seat basis, which is a beautifully simple and high-margin business model. But when your software becomes infrastructure, you start competing with the likes of AWS, GCP, and Azure, where the cost of goods sold is dominated by massive cloud computing requirements. Scott Brinker has noted that with AI becoming more powerful, many enterprises are looking at the 80% margins they pay to vendors and wondering if they could simply build their own applications directly on top of the cloud infrastructure, bypassing the vendor’s markup entirely.
This “build vs. buy” leverage is a nightmare for vendors that are mostly “UI and workflow” wrappers. If a SaaS platform doesn’t offer deep, defensible infrastructure that is too complex for a company to replicate—like a massive, pre-existing metadata layer or a global security fabric—its margins are toast. We are likely to see a massive churn in the martech ecosystem, where only the most “infrastructure-y” platforms survive. The companies that remain will have to fight for every penny of margin by proving that their “plumbing” is what makes the whole building livable, rather than just being a decorative facade that can be swapped out for a generic AI agent.
With the move toward consumption and outcome-based pricing, many are worried about the “tokenomics” of the industry. How should businesses navigate the risk of technology providers seeking a disproportionate share of the returns created by their agents?
The transition away from seat-based pricing to a “wallet” or “transaction” model is the most significant pricing experiment in the history of the software industry. We are currently in a period of heavy subsidization, where the cost of tokens is being kept artificially low by venture capital and bond markets funding massive data centers. Rudy Khoury from Fisher & Paykel hit on a vital point when he mentioned that consumption models are great for managing overheads in line with revenue cycles, but they carry a hidden danger. If you move to outcome-based pricing—where the vendor takes a cut of the value an agent creates—you risk handing over the keys to your profit margins.
The defense against this is what we call the “separation of the harness and the intelligence.” You must keep the control layer—the rules, permissions, and security—separate from the actual AI model. By doing this, you maintain “optionality.” If one provider tries to recover their losses by hiking token prices, you can pivot to an open-source model or a competitor like xAI without rebuilding your entire business logic. We are seeing businesses become very tactical about this, using tools like Salesforce Guardian or MuleSoft Agent Fabric to create a “harness” that prepares data and manages agents regardless of which specific AI model is doing the thinking. It’s about ensuring that the “intelligence” layer remains a commodity while you own the “harness” that connects it to your specific business measures.
There is a growing criticism of the “Hotel California” mentality regarding enterprise data—the idea that it is free to put data into a system, but you pay “through the nose” to get it out. How does the agentic era change the power dynamic between the customer and the data custodian?
The “Hotel California” problem is a classic pain point in the SaaS world, and it’s one that firms like Boomi and Informatica are highlighting more aggressively. For years, vendors encouraged you to dump all your customer information, workflows, and metadata into their clouds for free, only to charge you exorbitant fees for “outbound” data or sharing it with external systems. In the agentic era, data isn’t just sitting in a database; it’s being “activated.” Salesforce’s Data 360 pricing is a perfect example: they might let you ingest data for free, but the moment you want an AI agent to process that data and send it to a different destination, the meter starts running.
However, the power dynamic is shifting because agents need to be able to “see” and “act” across multiple platforms to be truly useful. If a vendor makes it too expensive or difficult to move data, they are effectively blinding the agents that their customers are increasingly relying on. This is why the concept of a “unified harness” is so important. Customers are starting to demand that their data, definitions of revenue, and churn metrics are portable and accessible by various “coworker” agents. If a SaaS provider keeps their data locked in a silo, they risk the customer moving their entire operation to a more open, infrastructure-centric platform that prioritizes interoperability over extraction fees.
As familiar logos and application screens disappear from the daily workflows of employees, can SaaS brands maintain the trust and loyalty they’ve built over decades, or does trust simply become a matter of “did the process work?”
This is a fascinating psychological shift for the corporate world. Liz Miller from Constellation Research has argued that we often mistake a familiar logo for the source of trust, but real trust is actually rooted in whether a system returns the right information and executes the process correctly. For a CMO, it shouldn’t matter if their team is working inside a branded dashboard or a blank Claude prompt, as long as the customer experience remains high-quality and consistent. We are essentially moving toward “headless” enterprise software. Just as headless CMS allows content to be delivered to any screen, these new AI-first interfaces allow business logic to be delivered through any chat window or voice assistant.
The loyalty of the future will be held by the Chief Information Officer, not the end-user. The CIO is the one who has to look at the underlying platform and decide if it can support the “skills” and “instructions” that agents need to follow. If an agent operating with a person’s privileges performs an unintended action or violates a security rule, that trust is shattered instantly. This is why Salesforce is leaning so heavily into products like “Guardian” and “MuleSoft Agent Fabric.” They know that if they can’t provide a safe, governed environment for agents to play in, the “invisible” trust they are trying to build will vanish. The logo might be gone, but the responsibility for the outcome is higher than ever.
Salesforce recently introduced “Agentforce” and the “Claude Force” integration, moving away from just providing a platform to providing ready-made agents like “Margot” for marketing. Does this suggest that the future of SaaS is more about selling “pre-trained labor” than selling tools?
That is exactly the direction the market is heading. The transition from tools to “coworkers” is a response to the fact that many customers have struggled to build their own effective agents from scratch. By offering ready-made agents for common sales, service, and marketing tasks, vendors are effectively selling outcomes rather than just features. When you deploy a marketing agent like Margot, you aren’t just buying a license to a piece of software; you are deploying a set of “skills”—instructions for following particular workflows and business measures.
This model, however, complicates the traditional “seat” commitment. We are seeing a move toward “flex credits” and “wallet” models where a customer might have fewer human seats in a contact center but requires more credits to power a fleet of digital agents. It’s a transition toward selling specialized digital labor. Salesforce’s internal pilot, which saw 4,000 daily active users interacting with these integrations within just a few days, shows a massive appetite for this. Whether it’s Deloitte, Siemens, or GitLab, the early adopters aren’t looking for better buttons to click; they are looking for agents that can authenticate with their credentials, respect access rules, and perform actions as if they were a highly trained employee.
What is your forecast for the martech ecosystem as it shifts further into the infrastructure layer?
I believe we are entering an era of “The Great Simplification” for the end-user, but “The Great Complication” for the architect. Over the next few years, the 15,000+ products currently cluttering the martech landscape will undergo a brutal consolidation. Only those that provide a unique, hard-to-replicate “infrastructure” value—such as a proprietary semantic layer or a specialized industry data model—will retain their margins. For the reader, the advice is clear: stop investing in “UI-first” tools and start building your “harness.” Focus on how your data is structured, how your metadata is defined, and how you can maintain the flexibility to swap out intelligence models as the “tokenomics” of the industry shift. The winners will be the companies that own their business logic and use SaaS as the plumbing to power an invisible, agentic workforce.
