A change in what software can do.
Conventional software follows instructions written in advance. Its behavior is described through logic: when this happens, do that. This remains essential. But it is no longer the only material we can build with.
Models made intelligence accessible through conversation. A person could describe a task, provide context, and receive an interpretation, a judgment, or a response.
Smart software takes the next step. Those capabilities become part of how the system operates. Software can interpret an event, assess its meaning, choose a path, and act within defined constraints.
Intelligence becomes part of the system architecture.
More than a conversational interface.
A conversation is one way to use intelligence. A system can also use it behind a workflow, inside an evaluation, or at a decision point. The interface may be familiar. The change is in what the software can understand and do.
This does not make every part of the system probabilistic. Explicit rules, typed interfaces, and deterministic code remain necessary. Intelligence extends the available operations; engineering defines where and how to use them.
New capabilities. New responsibilities.
When a system can make judgments, its context matters. When it can choose an action, its authority matters. When its behavior can vary, evaluation and feedback matter.
That is why smart software needs a discipline for the system around the model.
Explore engineered intelligence