In the race to build the next generation of enterprise software, a new battleground is emerging—not over who owns the data, but who owns the reasoning behind every business decision. As Jaya Gupta, Ashu Garg's recent article for Foundation Capital on context graphs puts it, “The UX of work is separating from the underlying data plane. Agents become the interface, but something still has to be canonical underneath.” The trillion-dollar question: what is that canonical layer, and how do we build it for the era of AI?
Why Data Alone Isn’t Enough
For decades, systems of record like Salesforce, Workday, and SAP have been the backbone of enterprise operations. They answer the question, “What happened?” But as AI agents begin to orchestrate more of our workflows, a new question becomes critical: “Why did we do that?” The answer isn’t found in a CRM field or a data warehouse snapshot. It lives in the messy, ad hoc world of exceptions, approvals, and cross-system context—often scattered across Slack threads, deal desk calls, and the collective memory of your team.
Foundation Capital calls this the “context graph”—a living record of decision traces, not just data points. But here’s the catch: most enterprises don’t have the structure or discipline to capture this context in a way that’s durable, queryable, and actionable. Retrofitting a context layer after the fact, based on what agents experience, is like trying to reconstruct a city’s traffic patterns from a handful of dashcam videos. You’ll get fragments, but never the full picture.
The Flaw in the “Retrofitting” Approach
The prevailing wisdom in much of the AI world is to let agents “learn” context as they go—scraping together decision traces from whatever digital exhaust they can find. In this approach you accept business operates as it does and you try to capture the decisions being made. But as our own research and experience at NOAN shows, this is fundamentally flawed. Why? Because enterprise knowledge is a mess: fragmented, inconsistent, and rarely treated as a first-class asset.
As I wrote recently, “Traditional AI approaches, such as Retrieval-Augmented Generation (RAG), struggle with disconnected knowledge, resulting in incomplete or misleading outputs.” The real problem isn’t just missing data—it’s missing structure. The reasoning that connects data to action is almost never captured as data in the first place.
I've lead transformation initiatives for global companies, and witnessed firsthand the chaos that results when organizations try to retrofit structure onto a foundation of fragmented, ad hoc knowledge. I have previously explored this in depth, you can read more in our FUBAR Mode article, the reality is that most enterprises operate in a state of “fragmented, unstructured, barely aligned reality”—where critical context lives in people’s heads, scattered chats, and disconnected systems. The research and real-world failures referenced in that piece make it clear: retrofitting a context layer after the fact is a fallacy. Without a deliberate, up-front commitment to fact control and structured knowledge, AI initiatives inevitably amplify confusion rather than clarity. My personal experience—and the research evidence—underscore why building a fact layer from the ground up is not just a technical preference, but a business imperative for any organization serious about AI-driven transformation.
The notion that AI agents can simply “unpick” the chaos of enterprise knowledge is built on a series of comforting—but deeply flawed—assumptions: that decisions are made rationally, that there’s a single, consistent workflow, and that employees always know what they’re doing. In reality, as revealed by a groundbreaking study of Fortune 500 teams, the truth is far messier. Participants described the moment they finally saw the full scope of their organization’s knowledge sprawl as an “epiphany”—one likened to the sun rising for the first time. The experience was so jarring that many left their roles to focus on organizational restructuring. One CEO, confronted with the findings, famously admitted, “This is even more fucked up than I imagined,” realizing his sense of control was pure illusion. I’ve witnessed this same reaction from C-suites at companies betting big on AI: the myth of order is only shattered when the full, tangled reality is finally exposed. Agents can’t reconstruct logic that never existed in the first place.
At NOAN, we took a true first principles approach—questioning not just how information is stored, but why. Most so-called “AI-native” solutions are still built on top of PDFs and documents, relics of the paper era whose formats haven’t changed since the quill. We asked: why do documents even exist? Their sole purpose is to transfer information, yet they do so in a format designed for parchment, not intelligence. The real mark of a first principles, AI-native company is how few documents it needs. In six months, the only PDFs we’ve created are legal necessities—and even those are on borrowed time. For us, it’s the facts themselves that transfer knowledge, not the containers. When you start from that premise, you don’t just digitize the past—you invent the future.
Fact Control: Building the Foundation for AI-Native Business
At NOAN, we believe the answer is what we call the fact layer: a unified, continuously updated source of truth that captures not just what happened, but why. Ultimately this is a context graph with a baked in single source of truth and integrated fact control. Fact control means collaboratively managing the lifecycle of facts—curating, verifying, updating, and contextualizing them—so that every stakeholder, human or AI, operates from the same foundation.
The fact layer is more than a memory bank—it’s an orchestration layer that structures and connects every piece of business knowledge into a single, living ontology. Each fact—whether it’s a strategic decision, compliance approach, an asset, or a detail about a lead, customer, or investor—is time-stamped, auditable, and interlinked within this unified system. This means every action and insight is not just stored, but contextualized and traceable, creating a transparent, AI-native system of record. The result is a business that can see, query, and trust the full lineage of its knowledge—no more silos, no more guesswork, just a continuously evolving foundation for smarter, faster decisions.
This isn’t a static database or a retrofitted log. It’s a living, dynamic system that evolves with every new insight, every exception, every precedent. As Foundation Capital notes, “If you persist those traces, you get something that doesn’t exist in most enterprises today: a queryable record of how decisions were made.” At NOAN, we make that record the starting point, not an afterthought.
Making the Invisible Visible
The fact layer is more than a technical upgrade—it’s actually a completely new way of running a business. By capturing the “why” behind every decision, we make the invisible visible and the ad hoc durable. This enables:
- Smarter, faster decision-making: Agents and humans alike can see not just what was done, but the reasoning and precedent behind it.
- Continuous learning: Every exception becomes a precedent, compounding organizational intelligence over time.
- Trust and auditability: When you can replay the state of the world at decision time, you can audit, debug, and improve with confidence.
Why NOAN’s Approach Is Different
Incumbents and data warehouses are trying to bolt on context after the fact, but they’re always a step behind. As Foundation Capital observes, “A system that only sees reads, after the fact, can’t be the system of record for decision lineage. It can tell you what happened, but it can’t tell you why.” NOAN sits in the orchestration path, capturing context at the moment of decision—making the “why” a first-class citizen in your business knowledge.
At NOAN, we believe the fact layer must be both human- and AI-controllable—designed so that people and intelligent agents can collaboratively create, update, and contextualize the knowledge that drives the business. This conviction is foundational to how we’ve built NOAN: as a platform where every decision, whether made by a human or an AI agent, is captured, structured, and made accessible as part of a living system of record. By enabling seamless, bi-directional control and transparency, NOAN ensures that the fact layer is always current, auditable, and actionable—empowering organizations to trust, adapt, and scale in an AI-native world.
As I highlighted in my recent article about the importance of Fact Control, “By adopting fact control, businesses can build trust, make informed decisions, and adapt to changes in real time, laying the foundation for AI-driven growth and innovation.”
When a company can present a living, queryable record of its business logic, performance, and decisions, it also transforms how it is perceived by external stakeholders. As Fintech Futures noted in their recent profile of NOAN, “Businesses can present their fact base directly to financial partners, enabling immediate understanding and decision-making. This transparency could unlock funding and investment opportunities for businesses that are typically underserved.” By making the “why” behind every number instantly accessible, organizations not only accelerate trust and due diligence but also open the door to faster partnerships, easier compliance, and a new level of business leverage. In this landscape, the fact layer isn’t just a technical upgrade—it’s a catalyst for growth and a competitive edge in securing capital and opportunity.
The fact based future of business
The future of enterprise AI won’t be built on bigger databases or smarter agents alone. It will inevitably be built on the fact layer: a living, queryable, and actionable record of business logic, decisions, and precedent. This is the missing link in enterprise AI—and NOAN is pioneering the way forward.
The greatest barrier for most businesses isn’t technical—it’s accepting that the leap to a fact layer or context graph is no longer optional. In a world where both agents and humans must operate from the same, living source of truth, clinging to legacy systems and fragmented knowledge will only widen the gap between the fast and the left-behind. The organizations that embrace this shift—structuring their knowledge as facts, accessible and actionable by all—will outpace, outlearn, and ultimately outlast those still patching together yesterday’s logic. The future belongs to those who build on facts; for everyone else, the distance to catch up will only grow.
In the era of AI, the companies that win won’t just know what happened. They’ll know why. And they’ll make that knowledge work for them, every single day.