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. The market is already delivering its verdict. Legacy enterprise SaaS companies are being systematically repriced as investors recognize that software built as shells for manual workflows will be obsolete. Not in a decade. Within years.
Why models aren’t enough
Large language models are rapidly becoming a commodity layer. This doesn’t mean that companies building LLMs are interchangeable. It means 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. 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 something else entirely: contextual understanding.
The context imperative
Context is what transforms vast datasets from noise into signal. An agent trained on enterprise data will only have historical outcomes from documented decisions. But human actors make decisions with information outside of what’s documented. They understand causality, relationships, and anomalies. They know what matters. That understanding, captured, encoded, and operationalized at scale, is context.
Context as foundation
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 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. Companies must now build the agentic enterprise on three interdependent requirements: context, reasoning, and agency.
Before we get any further, let’s define each of these terms in the context of this conversation:
- Context: The situational knowledge that helps AI understand relationships between pieces of information to produce appropriate outputs.
- Reasoning: The process by which an AI system uses logic or learned patters to reach conclusions and solve problems
- Agency: A system’s ability to independently make decisions and take actions to achieve goals
Critically, not all of these requirements are created equal. Context is the foundation for the other two requirements become valuable:
- 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 decision) is dangerous without context. An agent that acts without understanding what matters within a domain produces automated chaos at scale.
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.
The hierarchy is clear. Context is the necessary foundation the other requirements rest on top of. You can build sophisticated reasoning and extensive agency, but without context they won’t translate into durable business value. With it, they’re game-changing.
Building the context moat
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 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. Network-scale data across every major mode and geography. This isn’t a moat because it’s large. It’s a moat because it’s contextualized.
This video shows what the agentic enterprise actually looks like when context, reasoning, and agency work together. AI that doesn’t just suggest actions but executes them with the judgment of a veteran supply chain operator.
Can’t context be replicated?
The obvious objection: can’t competitors build similar context? In theory, yes. In practice, no. Not quickly enough to matter.
Building meaningful context requires three things most companies don’t have:
- Years of validated data collection across a network
- Deep domain expertise to interpret that data
- The operational workflows that prove the context actually drives better decisions.
This is not a data volume problem. It’s a data meaning problem. You can’t shortcut understanding what matters in supply chain logistics, healthcare, financial services, or any other complex domain. You earn that understanding through years of operational partnership with customers who trust you enough to share their most sensitive data.
By the time a competitor builds comparable context, the leaders will have compounded their advantage. Context delivers better AI outcomes, which attract more customers, which generate more data, which improve the context. Network effects occur at the data layer, not the user layer.
The payoff: AI that actually works
As context matures, it unlocks what AI can actually deliver: skills. Not chat interfaces or dashboards, but AI that does the work. Rerouting shipments before disruptions cascade. Resolving exceptions with the judgment of a 20-year veteran. Optimizing networks in real time.
Skills without context are parlor tricks. Every technology company will soon offer AI skills. The differentiation will come from the depth of context powering them.
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.
