All modelling - systems or otherwise - involves reduction. There is the global scale generalisation (to create laws, equations, concepts, etc) that follows from the shedding of the local particulars (this cat vs that cat, this mass vs that mass).
So reduction is the basis of modelling. The key difference is that reductionist reductionism

wants to reduce everything to a single kind of stuff (atoms, information, elements, components, materials) while the systems approach is dichotomistic and involves a reduction to both global and local scale boundaries (so atoms AND void, information AND meaning, substance AND form).
Systems approach also involves causality. Our logic is our model of causality (the reason why something has to happen). Again, ordinary thinking is monadic and so wants to reduce causality to local pushes and pulls only. Newtonian mechanics, computation, atomism, etc.
The systems approach says causality divides to become strongly local and also strongly global. So you have always both the something which stands for the local
constructive actions (such as the atomistic collisions, the vector forces, the computational steps), but also equally the causality represented by the system's global constraints (its organisation, its form, its 'emergent' laws and regularities).
So reductionism is good, as that is modelling, but there is the danger of over-reduction.
And systems causality is not really simple linear cause => effect as that is just a local scale description of causality (missing the global scale constraints which are also acting to create what is observed).
Again, Aristotle broke it into four causes, of which two relate to global constraint and two relate to local construction.
Aristotle just took it another step (probably unnecessary) in dichotomising the dichotomies.
He identified the
general local and global causes - the material and formal cause in his scheme (the general stuff of which it is made, the general form by which it is bound).
Then he also identified the
particular local and global causes.
He asked what was the particular local triggering event which caused something to happen (and this is about the only kind of cause that modern monadic, "cause and effect" modellingis concerned with - the non-system approach).
And Aristotle also asked about the particular global reason, purpose or meaning which also caused that thing to happen - the teleological level explanation.
Modern non-system thinking leads to a rejection of teleology as unscientific. But that is just a silly prejudice of course.
QM should have undermined it by now - cf: transactional interpretations, decoherence, non-locality. But there you go.