Using big-brain for target selection in an aiming system?

I'm figuring out how to model the following scenario: - An automated turret/laser needs to figure out which entity to target - Targetable entities have a value - The turret wants to maximise the total value of the enemies it's killing It seems like it should be straightforward - the turret is the `Thinker`, the action it can take is `Aim { target: Entity }`, and there is a `Score` for each targetable entity, derived from its value/worth. Questions: - Is this the right way to think about utility ai/big brain? It feels like my approach is problematic because there's no type-level difference between different Aim instances, so there's no 'decision' being modelled - If it can be done with big-brain, how can I ensure each targetable entity has a `Scorer`? How can I model actions which are too numerous to be known at compile time?