Vijay Raina has spent decades at the intersection of enterprise software and strategic transformation, making him a primary voice in the rapidly shifting world of SaaS and agentic workflows. As 2025 gave way to 2026, he witnessed a pivotal moment in technology where the first wave of AI agents moved from experimental demos to integrated enterprise tools, a transition that fundamentally altered how we perceive the “future of work.” From the early whispers of code-generating capabilities at Anthropic and OpenAI to the market-shaking “SaaSpocalypse” that forced Wall Street to reevaluate the value of software stocks, his perspective covers the high-stakes decisions today’s CIOs must make. In this discussion, we explore the cultural and operational ramifications of choosing between human-augmented AI and fully autonomous systems, the rise of the “context layer” as an organization’s secret weapon, and the technological pillars—from MCP servers to data fabrics—that are making it possible for companies like SAP to scale from a mere 40 agents in 2025 to more than 200 just a year later.
The industry seems divided between those who see AI as a tool to help humans and those who envision a completely autonomous enterprise. When leadership is sitting in the boardroom trying to decide which path to take, what factors should they weigh to ensure they aren’t just chasing a trend but building a sustainable operational strategy?
The decision between a fully autonomous “agentic” future and a human-augmentation model is not just a technical choice; it is a profound cultural statement that dictates how employees perceive their value within the company. When I look at the landscape today, I see companies like Atlassian leaning into the “human-AI collaboration” narrative, while others like SAP are aggressively pushing the concept of the “autonomous enterprise.” CIOs must understand that for many, human-in-the-middle is a necessary transitional phase designed to build the deep trust required for AI to handle high-stakes decision-making. Recent data highlights the urgency of this transition, with a Deloitte report showing that 36% of IT leaders expect at least 10% of their jobs to be fully automated within a single year, a number that jumps to a staggering 82% within a three-year window. Leadership must evaluate whether they are deploying AI in operationally critical or customer-facing domains where a human touch is essential for innovation and critical thinking, or in areas where reliability at scale is the only priority. Ultimately, the way a leader positions these agents will determine whether they face vocal detractors fueled by job-loss fears or an empowered workforce that views AI as a partner in productivity.
We recently witnessed a significant selloff in SaaS stocks that many are calling the “SaaSpocalypse,” driven by the fear that AI code generators might allow companies to replace third-party software with in-house builds. How should an organization navigate the “build versus buy” dilemma without falling into the trap of accumulating massive AI debt?
The “SaaSpocalypse” was a visceral reaction from Wall Street investors who feared that advancements from OpenAI, Anthropic, and tools like Replit or Claude would allow CIOs to bypass traditional SaaS vendors entirely. While it is true that code generators are now building applications that The New York Times describes as “credible,” even if they remain occasionally flawed, the reality is that writing code is only one small fragment of the development lifecycle. Organizations that jump too aggressively into building their own agents often ignore the massive overhead of security, governance, and long-term maintenance, leading to what we now call AI debt. I often advise engineering teams to look at the tools provided by established players—such as Atlassian Rovo Dev, Pega Infinity Studio, or Snowflake CoCo—because these platforms already have the necessary data infrastructure and governance “baked in.” By choosing these integrated development environments instead of a pure do-it-yourself approach, companies can avoid the hidden costs of managing custom-coded agents and instead focus on technical proofs that validate their data and modeling capabilities.
The growth in the number of available AI agents has been explosive, with some vendors increasing their offerings fivefold in just twelve months. What are the specific technological engines driving this rapid expansion, and how can a CIO possibly keep up with the evaluation of so many new tools?
The acceleration we are seeing is truly unprecedented; consider SAP, which scaled from 40 Joule Agents in 2025 to over 200 in 2026, a clear signal that the floodgates have opened. This growth is being fueled by three specific technological pillars: the maturation of data fabrics from companies like Salesforce and Boomi, the adoption of MCP servers for agent-to-agent communication, and the launch of proprietary agent development tools like Appian Composer and Adobe Firefly AI Assistant. These advancements mean that agents are no longer isolated bots but are part of a multi-step agentic workflow that can navigate complex business processes across different platforms. For a CIO, this creates a massive management challenge that requires dedicated analysts to weigh the trade-offs in cost, compliance, and performance. To avoid the risk of “shadow AI” or employee confusion, organizations must implement a transparent, defined process for selecting, reviewing, and monitoring these agents, ensuring that every deployment is backed by solid end-user feedback and a clear ROI.
You often speak about the “context layer” being the real differentiator for AI in the enterprise. Could you explain what this layer actually consists of and why a company’s internal data is more important than the underlying AI model itself?
The context layer is essentially the “intelligence bridge” that sits between the raw AI models and the structured reality of an enterprise, acting as the knowledge base that tells an agent how to act in a specific business environment. It is composed of semantic layers, knowledge graphs like the Atlassian Teamwork Graph or the SAP Knowledge Graph, and cleansed document repositories that provide the necessary signals for an agent to make a recommendation. While tech vendors will always compete on the raw power of their LLMs, the true “secret sauce” for any company is its own trusted data and well-defined business processes which are unique and cannot be replicated by a generic model. We are seeing a new wave of governance tools, such as the Snowflake Horizon Catalog and the Quickbase AI Control Center, specifically designed to manage this context layer at scale. If an organization does not have a solid handle on its data catalog and internal guardrails, even the most sophisticated AI agent will fail because it lacks the “human-to-agent” and “agent-to-agent” context required to perform meaningful work.
We are moving away from traditional static dashboards and forms toward what you call “conversational user experiences.” How does this shift change the daily life of an average employee, and what should change management programs focus on during this transition?
The shift toward conversational UX marks the end of the era where employees had to spend their days wrestling with static reports, complex flows, and rigid forms. Instead, we are entering an era of “coworkers” and “assistants,” where platforms like Adobe CX Coworker or Nutanix NIVA allow marketers and engineers to interact with their tools using natural language and “vibe coding.” These AI-first user experiences are designed to augment human capability, managing campaigns or monitoring performance through simple prompts rather than manual data entry. For employees, this feels less like operating software and more like collaborating with a teammate who has instant access to every piece of company data. CIOs must ramp up their change management programs to focus on this new reality, helping workers move past the fear of automation and instead embrace the productivity improvements these assistants offer. The goal is to drive adoption by demonstrating how tools like Appian AI Copilot or Cisco AI Assistant can handle the drudgery, leaving the creative and strategic work to the humans.
What is your forecast for the evolution of AI agents over the next two years as they move beyond simple productivity gains?
I predict that we are about to move past the initial “efficiency phase” of AI and into a period where agentic workflows become the primary engine for top-line business growth. While the current focus is heavily on saving time and automating repetitive tasks, by the end of 2027, we will see the first true wave of AI-driven digital transformation where agents are used to evolve business models and create entirely new customer-facing products. We will see a consolidation of the “context layer” where the various knowledge hubs and semantic graphs from major vendors begin to interoperate more fluidly, allowing a Snowflake agent to seamlessly hand off a task to an SAP agent without human intervention. This will lead to a new standard of “autonomous growth,” where AI doesn’t just help us do our current jobs faster but identifies new market opportunities and executes on them in real-time. The organizations that thrive will be the ones that have already spent the last year building the data foundations and governance guardrails necessary to let these agents run at full speed.
