Re policy, it's used e.g. in environmental resource management (regional level), anti-terrorism (networks, opinion dynamics), urban planning, disease control.
Policy isn't based alone on it, of course. But it's used as an ingredient. Often the models aren't good enough (lack of data and quantifiable causal relationships) to gain an exact image of the modeled system, but they provide a general system understanding (system behavior under different [intervention] scenarios) that is often educational and gives insights that then need to be empirically validated.
A barrier to increase its use in policy-making is the "science-policy gap" between researchers and decision-makers. Either they don't understand the utility of model results and dismiss them or they trust them too much--both is not good. So researchers had to come up with ways to communicate results, often include decision-makers (and other stakeholders) in the modeling process.
Depending on how you classify him (as an economist or a sociologist), Thomas Schelling's work on residential segregation has likely informed policy over the last decades.
Recently a nice animation of his dynamic models of segregation processes was posted here: http://ncase.me/polygons/
I mean, the field has inspired how Facebook determines to serve up news feed content so, in a way, it has driven policy by way of influencing democracy.
A barrier to increase its use in policy-making is the "science-policy gap" between researchers and decision-makers. Either they don't understand the utility of model results and dismiss them or they trust them too much--both is not good. So researchers had to come up with ways to communicate results, often include decision-makers (and other stakeholders) in the modeling process.