BSFN Compositions

Ask what an item’s availability at a branch is and you can read a number from a table — and be wrong, because availability is not a column. It is a computation: on-hand, less commitments, less holds, plus whatever your configuration says counts. JD Edwards encodes thousands of such rules in its business functions, refined over decades and tuned by your own setup.

A BSFN Composition packages one of those functions as a capability. Ask for the availability and you get the number JD Edwards itself would compute — respecting the same rules, the same edits, the same configuration — not an approximation reconstructed from raw tables.

One capability, one action

A BSFN Composition is a single, well-defined business action: validate this combination, compute this figure, look this up the way the application would, perform this master-data operation. The author names it, describes what it needs and what it returns in business language, and that description is exactly what the reviewer approves and the AI receives. Nothing about it is open-ended.

Deterministic where it matters

Agentic AI is at its best when the judgement is the model’s and the arithmetic is not. BSFN Compositions put the deterministic part of a task — the validation, the calculation, the rule — behind a capability that behaves identically every time, so the AI reasons about the business while JD Edwards does the accounting.

Governed like everything else

Like every Composition, a BSFN capability is reviewed before publication, published to named groups, visible only to entitled users, and withdrawn centrally in one act. What your AI can do remains a list a person signed off — now with JD Edwards’ own logic on it.