Agricultural intensification (AI) is currently a major driver of biodiversity loss and related ecosystem functioning decline. However, spatio-temporal changes in community structure induced by AI, and their relation to ecosystem functioning, remain largely unexplored. Here, we analysed 16 quantitative cereal aphid-parasitoid and parasitoid-hyperparasitoid food webs, replicated four times during the season, under contrasting AI regimes (organic farming in complex landscapes vs. conventional farming in simple landscapes). High AI increased food web complexity but also temporal variability in aphid-parasitoid food webs and in the dominant parasitoid species identity. Enhanced complexity and variability appeared to be controlled bottom-up by changes in aphid dominance structure and evenness. Contrary to the common expectations of positive biodiversity-ecosystem functioning relationships, community complexity (food-web complexity, species richness and evenness) was negatively related to primary parasitism rates. However, this relationship was positive for secondary parasitoids. Despite differences in community structures among different trophic levels, ecosystem services (parasitism rates) and disservices (aphid abundances and hyperparasitism rates) were always higher in fields with low AI. Hence, community structure and ecosystem functioning appear to be differently influenced by AI, and change differently over time and among trophic levels. In conclusion, intensified agriculture can support diverse albeit highly variable parasitoid-host communities, but ecosystem functioning might not be easy to predict from observed changes in community structure and composition.