Environmental and agricultural policy instruments cause changes in land-use which in turn affect habitat quality and availability for a range of species. These policies often have wildlife or biodiversity goals, but in many cases they are ineffective. The low effectiveness and the emergence of unwanted side effects of environmental and agricultural policies are caused by over-simplistic assumptions in the design of policy instruments as well as difficulties with predicting behaviours of policy subjects. When considering wildlife in agricultural landscapes, policy’s performance depends both on human (farmers) actions, which the policies aim to affect, and wildlife responses to land-use and management changes imposed by farmers. Thus, in order to design effective agri-environmental policies, detailed ex-ante assessments of both of these aspects are necessary. Due to the restrictive assumptions and technical limitations, traditional agricultural economic and ecological models fall short in terms of predictions of impacts of agri-environmental measures. The feedback situation between policy, human behaviour and ecological systems behaviour can confound these approaches, which do not take systems complexity into account. Therefore, a solution that integrates both feedback interactions and the differing scales at which these interactions take place is needed. For this, we suggest developing integrated policy assessment tools comprising of simulated farmer decision making, on-farm land-use and wildlife responses in the form of spatially explicit, dynamically connected agent-based models. Although complex and necessitating true inter-disciplinarity, these approaches have matured to the point where this endeavour is now feasible.