This paper analyses the integration of active labour market policies for two groups of unemployed from a theoretical perspective. In general a model with only one type of agent performs better than a model with two types of agents. If there are two types of agents part of the effort of one agent leaks away to the other agent and decreases the incentives to get the unemployed back to work. A model where two agents work together and serve both types of unemployed performs even worse. This is because they are only partially compensated for their effort, which decreases the incentives to get the unemployed back to work even more.