This paper presents a novel application of agent-based simulation software to tune real greenhouse infrastructure containing flowering seed or vegetable crop plants and their insect pollinators. Greenhouses provide controlled environments for the growth of high-value crops. As global climate and weather become more unpredictable, we are becoming more dependent upon technologically sophisticated greenhouses for reliable crop production. For crop pollination in a greenhouse, although manual or technological alternatives have been explored, pollination by bees is still required in many crops for the best seed yields and food quality. However, the design of greenhouses is driven primarily by the requirements of the plants rather than the pollinators. In light of this, we have designed simulations to explore improvements to greenhouse conditions and layout that benefit the insect pollinators and assist them to pollinate the crop. The software consists of an agent-based model of insect behaviour that is used to predict pollination outcomes under a range of conditions. The best parameters discovered in simulation can be used to adjust real greenhouse layouts. We present a key test case for our method, and discuss future work in which the technique has the potential to be applied in a continuous feedback loop providing predictions of greenhouse re-configurations that can be made by real-time control systems in a modern greenhouse. This is a novel approach linking simulation behaviour to real techno-ecological systems to improve crop and seed yield from valuable greenhouse infrastructure.