Vijay Raina is a seasoned authority in the realm of enterprise SaaS technology, bringing years of experience in software design and architecture to the forefront of the industry’s most pressing debates. As organizations navigate the transition from traditional cloud services to AI-driven ecosystems, his insights help bridge the gap between technical infrastructure and strategic business value. In this discussion, we explore the shifting landscape of enterprise software, examining whether the rise of autonomous agents signifies a “SaaS-pocalypse” or a historic expansion of opportunity for platforms that can successfully integrate these new capabilities into the core of the modern workflow.
Transitioning from per-seat pricing to usage-based metrics like Agentic Work Units is a major shift. How does this change the way enterprise revenue is forecasted, and what specific steps should software companies take to ensure these metrics accurately reflect the outcomes generated by AI?
The shift toward Agentic Work Units fundamentally alters the financial rhythm of a company because it moves the focus from human headcount to the tangible value delivered by machine intelligence. In the old world, we looked at how many employees were using a tool, but now we must forecast based on the volume of “outcomes” or specific tasks successfully completed by AI agents. To ensure these metrics are accurate, companies must first establish a granular baseline for what constitutes a “unit” of work, such as a resolved support ticket or a completed sales sequence, and then implement a step-by-step tracking system that differentiates between simple automation and high-value decision-making. By pivoting away from selling mere access and toward selling results, firms can stabilize their revenue pipelines even as the number of human “seats” fluctuates. This requires a rigorous audit of existing pipelines to see where AI is actually doing the heavy lifting, ensuring the billing is tied directly to the productivity gains the customer experiences in real-time.
As AI agents automate routine customer service tasks, the traditional subscription model faces pressure from a potentially shrinking human user base. What strategies can platforms use to maintain their value proposition, and can you share an anecdote where automation actually increased a client’s platform engagement?
To maintain a strong value proposition in an era of shrinking human seats, platforms must reposition themselves as the essential “operating system” for AI-driven operations rather than just a tool for human hands. The strategy here is to deepen the integration into the customer’s core business data, making the platform the “brain” that orchestrates various AI agents. I’ve seen cases where a customer service department used AI to automate nearly 80% of their routine queries, which initially led to fears of reduced platform usage. However, the exact opposite happened: because the routine tasks were handled, the remaining human staff began using the platform more intensely to tackle complex, high-value customer issues they previously didn’t have time for. This increased the overall engagement with the platform’s advanced analytics and data features, proving that while the “seat count” might change, the depth of the interaction often grows much richer and more vital to the business.
While some believe generative tools allow businesses to build their own lightweight applications, enterprise systems still require deep integration and security compliance. Why is it difficult for standalone AI to replicate these complex workflows, and what specific data preparation challenges must organizations overcome first?
Standalone AI models often lack the “connective tissue” required to operate safely and effectively within a large-scale corporate environment. An enterprise system is not just a collection of code; it is a fortress of security, compliance, and pre-integrated data flows that have been refined over decades. For a standalone AI to replicate this, it would need to manage complex permissions and regulatory requirements that go far beyond simple natural language processing. Before an organization can even think about replacing these systems, they face a massive data preparation hurdle where they must clean, structure, and secure their internal information so that an AI can actually understand it without causing a security breach. This “data readiness” is a significant barrier that makes the deep integration of established platforms like Salesforce more attractive than trying to build a fragmented, home-grown alternative from scratch.
There is growing concern that value in the technology stack is moving away from application layers toward AI model providers and infrastructure. How can established platforms avoid being reduced to back-end systems, and what specific indicators suggest a company is successfully retaining its pricing leverage?
The danger for application-layer companies is becoming a “dumb pipe” for more powerful AI models, but they can avoid this by owning the user interface and the workflow context. If a platform controls the specific environment where a business decision is made—such as the final approval of a contract or the management of a customer relationship—it retains its status as the primary interface. A key indicator that a company is successfully retaining its pricing leverage is the stability of its margins even as it integrates third-party AI models. If a firm can introduce usage-based pricing for AI agents and see its average revenue per account grow despite a lower seat count, it shows that the customer values the platform’s orchestration of the AI more than the underlying model itself. Furthermore, seeing a healthy growth in the sales pipeline for AI-embedded products suggests that the market still views the application layer as the place where real business value is captured and realized.
Embedding AI has shown productivity gains in routine queries, yet the technology often struggles with nuanced, high-stakes interactions. In what ways can AI complement rather than replace human workers, and what step-by-step training is necessary to bridge the gap between AI promise and financial reliability?
AI is at its best when it acts as an “augmented intelligence” that strips away the drudgery of routine tasks, allowing human workers to focus their unique emotional and creative talents on high-stakes interactions. To bridge the gap between the promise of AI and real-world financial reliability, organizations should follow a structured three-step training approach. First, deploy AI in a “shadow” mode where it suggests answers to humans for review, allowing the system to learn from expert feedback without any risk to the customer experience. Second, move to a “supervised automation” phase where the AI handles low-risk routine queries but flags anything nuanced or emotionally charged for immediate human intervention. Finally, conduct continuous data audits to ensure the AI remains aligned with the firm’s specific enterprise trust and compliance standards. This phased rollout ensures that the technology builds a foundation of reliability, proving its worth in dollars and cents before it is given more autonomy.
What is your forecast for enterprise software?
I believe enterprise software is entering a period of massive consolidation where “essential infrastructure” platforms will thrive by becoming the orchestrators of AI, rather than just the hosts of human users. While the transition from per-seat pricing will create some short-term turbulence and uncertainty in the markets, the long-term opportunity for platforms that can successfully manage enterprise-grade data and complex workflows is actually expanding. We will see a shift where the value is no longer in how many people use the software, but in how much “work” the software itself performs autonomously. Ultimately, those companies that can bridge the gap between raw AI models and the strict requirements of security, trust, and integration will not only survive this transition but will become more deeply embedded in the corporate stack than ever before.
