The verdict, and the smallest fix
Before a Relationship Manager sends a proposed rebalance, Luscent checks every trade against the client's mandate and the strictest jurisdiction that applies, and returns a verdict, a plain reason, and if it blocks, the smallest change that would clear it.
This is the core of the product, and it went in this month.
Before a Relationship Manager sends a client a proposed rebalance, the Suitability Gate checks every trade in it against that client's mandate and the strictest jurisdiction that applies. It returns one of three verdicts: recommend, escalate, or block. With the verdict comes a plain reason a person can read, and when it blocks, the counterfactual: the smallest change to the proposal that would clear it.
No model casts this vote. The gate is deterministic, so the same proposal always gets the same answer, and the reasoning is written out rather than inferred. An RM can override, and the override is recorded with its reason. Every verdict and every override is written to the Audit Trail, so a compliance officer can reconstruct why any decision was made, months later.
Nothing here moves money. It is a gate on advice, before send. It is what the EU AI Act asks for under Article 14, human oversight of an automated decision, built as a control rather than a promise.