For three years the strategy conversation in most enterprises has been a conversation about models. Which foundation model. Which vendor. Which benchmark, which context window, which fine-tune. The assumption underneath all of it was that the model was the asset, and that owning the best model - or the best access to one - would translate into durable advantage.
It will not. The model was never the moat.
Capability is converging, access is commoditizing
The uncomfortable fact is that frontier capability diffuses fast. A model that felt like a competitive secret in one quarter is matched by an open-weights release two quarters later, and the gap between the best proprietary model and the best available-to-everyone model keeps compressing. Whatever your competitor cannot do today with a model, they will be able to do next year with a cheaper one.
If the capability is purchasable, it is not a moat. A moat is something your competitor cannot replicate by signing the same contract you signed. And every enterprise can sign the same contract.
So the strategic question is not which model but what does the model have access to that no one else can assemble. That is a question about context.
Context is the part that does not commoditize
By context I do not mean the prompt, and I do not mean a pile of documents in a vector store. I mean the full situated knowledge a decision requires: the policies that constrain it, the precedents that inform it, the entities involved and their histories, the obligations in force, the reasons prior decisions went the way they did. Most of that is not written down anywhere a model can reach. It lives in people’s heads, in closed channels, in the gap between what a document says and what the organization actually does.
That gap is the moat. It is expensive to assemble, specific to your business, and impossible to buy. The work of turning it into something a system can use - capturing it, structuring it, keeping it current, and attaching governance to it - is the work that actually compounds.
A model with generic context produces generic answers with great fluency. The fluency hides the genericness, which is precisely what makes it dangerous in an enterprise setting.
Why “governed” is the load-bearing word
It is not enough for context to be present. In an enterprise, context has to be governed - meaning every piece of it carries provenance, permissions, freshness, and accountability. A retrieved fact that no one can vouch for is not an asset; it is a liability waiting for an audit. The difference between a demo and a production system is almost entirely the difference between ungoverned and governed context.
This is where the context intelligence framework starts: treating context as infrastructure with the same rigor you would apply to a database or an identity system. Who can see this context. When was it last true. What decision did it inform. What is the chain of custody from source to answer. These questions are unglamorous and they are the entire game.
The architecture that operationalizes this - what I call the Context Intelligence Infrastructure - is not a feature bolted onto a model. It is the layer that decides what the model is allowed to know, on what authority, and with what record. The model is a tenant in that house, not the house.
The strategic inversion
Once you accept that context is the asset, your investment priorities invert. You stop chasing the next model release as though it were a strategy. You start asking how much of your organization’s situated knowledge is legible to a system at all, and how much of it decays silently because no one captured it at the moment it was true.
You begin to treat the model as interchangeable - because it is - and the context layer as the thing you are actually building a company around. When the next, better, cheaper model arrives, you swap it in behind a context layer that does not change, and you keep your advantage. Your competitor, who built around the model, has to rebuild.
The teams that understand this are quietly building the boring infrastructure that does not demo well: provenance, permissioning, capture, decay management. The teams that do not are still in the model-selection meeting, certain that this quarter’s benchmark winner is the one that finally matters.
The model was never the moat. The context was, the governance was, and the discipline to treat both as infrastructure rather than afterthought is what will separate the enterprises that compound from the ones that keep starting over.