The conversation around enterprise technology has entered a decisive phase. Across industries, leaders are weighing the impact of emerging tools, questioning how automation and intelligence can genuinely transform their businesses rather than just add more pilots to the stack. In his session “Systemic AI: The Next Paradigm in Enterprise Transformation” at MachineCon 2025, Anuj Krishna, Co-Founder and President of Technology & Growth at MathCo, placed this discussion into sharper perspective. He argued that while enterprises have experimented widely with new technologies, what is required now is a systemic approach that brings purpose and context together. He called this shift “Systemic AI”.
Krishna began with an anecdote that captured the mood of the executives he meets. A pharmaceutical CEO told him, half in jest and half in frustration, that if another visitor mentioned the word “agentic” in his office, the meeting would end abruptly. This represented a broader AI fatigue in the market. Over the past two years, organizations have seen a flood of proof-of-concepts around generative AI, and more recently, agentic AI, yet many of these experiments remain isolated. As Krishna explained, leaders are asking whether these pilots truly deliver business outcomes. “We have seen a lot of POCs with respect to agentic and generative AI. We have also seen questions around the ROI that people are getting out of it. There has been some disillusionment with how useful it really is,” he said.
The limitations, particularly that of agentic AI, he argued, are tied to how the conversation has been framed. Much of the focus has been on the idea of agents, with insufficient attention to what they are meant to accomplish. “Agentic AI emphasizes the word agent and does not talk enough about what the agent is going to do,” he observed. This emphasis produces implementations that may look sophisticated on dashboards but struggle to integrate with core systems. The result is context blindness, technically correct answers that do not translate into actionable insights, and siloed solutions that never scale across the enterprise.
Krishna’s alternative is a pivot toward systemic thinking. Systemic AI, as he defined it, is about embedding purpose and context at the center of enterprise transformation. The shift is from focusing on the capabilities of individual agents to asking what business outcomes they serve, what processes they fit into, and how they orchestrate across functions. The distinction is subtle but powerful: while the agentic approach builds isolated assistants, the systemic approach weaves intelligence into workflows in a way that reflects organizational knowledge and objectives.
To illustrate the principle, Krishna introduced a framework with two axes. On one axis lies purpose, which grows in complexity from simple to advanced. On the other lies context, which deepens as tasks demand greater organizational knowledge. At the lower left, where purpose is simple and context is light, sit tasks such as document summarization or text-to-SQL conversion. Moving upward, tasks like market research require a richer context to interpret business goals. Higher still, sales or marketing copilots need awareness of user roles, objectives, and organizational dynamics. At the top, broad initiatives such as reducing churn or improving marketing ROI demand the highest combination of complex purpose and deep context. “When you stray off this path and try to achieve complex objectives using low context, you are not going to get value,” Krishna cautioned.
The implications for enterprise strategy are significant. Many organizations are tempted by the flood of new tools, each promising efficiency gains in a particular area. Krishna urged caution. Individual productivity tools or narrowly scoped platforms can be useful, but real transformation requires workflows that cut across departments. “I have seen multiple organizations where there are deep silos between one function and the other,” he said. “You cannot break them unless you bring them together in some way, and for that, you need the context sitting in either place.” The goal is not to add tools for their own sake but to align them with organizational context and orchestrate them for measurable outcomes.
Krishna outlined what such an architecture could look like, describing four core building blocks that power Systemic AI and form its foundation:
Foundation Layer – This sits at the bottom of the architecture and is the most critical, yet the least commoditized. It is primarily enterprise-owned and comprises data, metadata, workflows, and processes — essentially all the data and context required to build AI solutions.
Intelligence Layer – This hosts models and agents, some of which can be commoditized or enterprise-owned. What truly differentiates this layer is the orchestrator, which brings coherence and synergy across models and agents.
Value Layer – This layer focuses on adoption and impact. It houses use cases, applications, and interfaces that directly connect with humans, ensuring that AI delivers tangible business outcomes.
Strategy & Governance Layer – Cutting across all other layers, this ensures that the architecture remains aligned with company strategy and goals while maintaining governance and accountability at all times. “Think of governance at a block level,” he advised. “The way you think of governance for the context layer versus how you think of governance for agents versus how you think of governance for the interface will be different and needs to be different.”
Throughout his talk, Krishna returned to the theme of context as the true enterprise asset. “Enterprises need to own their own context and semantic layer. That is yours. That is your IP. Nobody else is going to give it to you,” he stressed. While external platforms may offer a starting point, the responsibility for embedding organizational processes, data, and goals rests with the enterprise itself. Context, in his view, is what makes AI systemic rather than superficial, and what differentiates one organization’s transformation from another’s.
Krishna also addressed how to begin the journey. He recommended starting small, with a “slice” of the business. This might mean focusing on a single function such as marketing or even a sub-function like primary market research. Building systemic AI slice by slice allows organizations to refine feedback mechanisms, establish governance, and build trust before scaling across larger initiatives. Trust, in particular, requires attention to user experience. “We are still far away from a world where it is completely autonomous. Until that happens, you need to bring the know-how of users into the system, and the only way to do that is to make them comfortable with it, make them trust it, and make sure that feedback is thought about in a good way,” he said.
As he closed the session, Krishna summarized the priorities that enterprises should carry forward. Organizations must claim ownership of their context and semantic layer as core intellectual property. They should approach transformation one slice at a time, scaling with discipline rather than breadth. They must center users in the design of systems, ensuring adoption and trust. Finally, governance must be designed distinctly for each building block of the systemic architecture rather than applied as a blanket framework.
