The AI revolution isn’t about the models. It’s about what we build on top of them.
This insight, articulated by Salesforce founder Marc Benioff in his recent TIME piece ahead of Davos 2026, crystallizes a truth that should fundamentally reshape how enterprise technology companies think about their competitive advantage. To quote Marc directly, “what we build on top of the LLM—the trusted data and workflows that connect AI to the way we work and live—matters most.”
For those building enterprise technology, this thesis carries existential weight. And the market is already delivering an initial verdict: legacy enterprise SaaS companies are being systematically repriced as investors recognize that software built as shells for manual workflows will be obsolete within years, if not sooner. The last 24 hours on Wall Street has furthered this argument. Legacy business models need to advance to survive.
The writing is on the wall. The agentic enterprise will look different and so will the interface between humans and software. What’s missing from the broader conversation is what is needed to build software for this new age. AI agents and LLMs alone are insufficient for managing the complexity of many global enterprises.
Large language models are rapidly becoming a commodity layer. This does not mean that companies building LLMs are interchangeable, but it does mean that durable differentiation no longer comes from the model itself. The advantage shifts to what sits above the model: trusted data, deeply embedded workflows, and domain-specific understanding that connects intelligence to real operational decisions
As models become more powerful and accessible, a paradox emerges. The more capable the models, the more valuable the unique, network-scale data and deep industry context powering that model becomes.
Context Is The New Moat
Context is what transforms vast datasets from noise into signals. It’s the ability to understand what matters within a specific domain, to filter the irrelevant, and to surface insights that drive the correct action. An agent trained on the data of an enterprise will only have data from that enterprise and historical outcomes from previous decisions of that enterprise. But human actors make decisions with information outside of the decisions documented from their work.
Every enterprise technology company will soon have access to powerful AI models. Many will have large datasets. But the companies that win will be those that have spent years building contextual understanding: the domain expertise, the relationship mapping, the anomaly detection, and causal modeling that makes their data uniquely valuable for decision-making. At project44, we’ve spent years building the world’s largest logistics data graph specifically to enable this agentic future. Over 250,000 carriers connected, 1.5+ billion shipments tracked annually, 7.3 trillion data points ingested and validated annually, and network-scale data across every major mode and geography.
Without context, enterprise AI deployments will fail. MIT research has already shown that 80%+ of AI projects have struggled to achieve meaningful outcomes with half of those projects failing due to context and data gaps. However, as context is introduced, hallucinations fall, and AI becomes both more effective and efficient.
Context as the Foundation for Agency and Reasoning
Traditionally, we’ve thought about platform development as a linear journey: first connect the data sources, then add context, then enable actions, finally automate workflows. This sequential model made sense when each capability built on the previous one and when development cycles could stretch across years.
But we don’t have years anymore. The rapid devaluation of legacy SaaS proves that markets move faster than traditional roadmaps. The agentic enterprise rests on three requirements: context, reasoning, and agency. These requirements must work interdependently and simultaneously, reinforcing one another rather than being developed as a linear sequence.
However, not all requirements are created equal. Context is the foundational requirement upon which the others depend.
Take exception recovery as an example. A predicted delay on an ocean vessel does not have a single correct response. In one case, it may trigger a downstream inventory reallocation to protect customer service commitments for a top-tier account. In another, it may require no action at all because safety stock and downstream capacity absorb the disruption. The data is identical. The outcome is not. Context determines the difference.
And while reasoning and agency are needed for agentic systems, they need context. Reasoning (applying AI models and logic) is powerful but directionless without information to reason with. Even the most sophisticated language models cannot determine what is important within a specific domain without a contextual framework to guide interpretation.
Agency (taking action and automating decisions) is dangerous without context. An agent that acts without understanding what matters within a domain produces automated chaos at scale.
The Hierarchy Is Clear
These four requirements don’t work sequentially, but they do have a hierarchy: context is the foundation upon which the others become valuable. You can build sophisticated interoperability, powerful reasoning, and extensive agency, but without context they won’t translate into durable business value. With it, they are game-changing.
The companies winning in the agentic era will be those that recognized this hierarchy early and built context first and deeply, then layer the other requirements on top of it.
Deterministic Workflows in the Global Supply Chain
Whether a company makes and sells cars, clothing, or drugs, there are explicit operational workflows tied to planning, making, and moving goods. These workflows (if x then y) are deterministic in nature. And they are deterministic because the cost of diverging from the critical path in the wrong way can cost companies hundreds of millions of dollars.
As we look ahead to where AI can play a role in reshaping the supply chain, it can complement and even replace deterministic workflows. But the bar is high to being better and/or faster than determinism. Context will be mission critical to achieving that outcome for enterprise use-cases.
The Next Horizon is Here
The agentic enterprise isn’t a future state. It’s happening now. Without context, the entire premise falls apart. But with it, enterprise software companies will navigate this monumental shift, and AI-native companies will be able to have the impact on humanity that we think they can.
