Function calling
The mechanism by which a model produces a structured request to call a named function with typed arguments, which the surrounding system executes.
In more detail
Function calling is the concrete implementation behind most tool use. Rather than asking the model to write free-form instructions, the system gives it a set of function signatures, and the model responds with a structured call: the function name and the arguments, formatted so software can execute it directly. The result flows back to the model to continue.
The structure is what makes it reliable. Typed arguments can be validated before execution, unknown functions can be rejected, and the whole exchange is machine-readable rather than parsed from prose. Function calling is effectively structured output aimed at triggering actions.
Where this shows up at Ceven
Under the hood, Ceven's AI steps invoke connectors through structured, typed calls rather than brittle text parsing, which is what lets a model's decision reliably become a real action across the customer's tools. Those calls are validated, gated where consequential, and recorded, so the path from decision to action stays legible.