Standard micro-economics concentrate on the description of markets but is seldom interested in production. Several economists discussed the concept of a firm, as opposed to an open labour market where entrepreneurs would recrute workers on the occasion of each business opportunity. Coase cite{Coase} is one of them, who explains the existence of firms as institution because they reduce the transaction costs with respect to an open labour market. Whatever the rationale proposed by economists to account for the existence of firms, their perspective is based on efficiency and cost analysis. Little attention is paid to the dynamics of emergence and evolution of firms. The aim of the present manuscript is to check the global dynamical properties of a very simple model based on bounded rationality and reinforcement learning. Workers and managers are localised on a lattice and they choose collaborators on the basis of the success of previous work relations. The choice algorithm is largely inspired rom the observation and modeling of long term customer/sellers relationships observed on perishable goods markets discussed in Weisbuch etalcite{Weisbuch} and Nadal etalcite{Nadal}. The model presented here is in no way an alternative to Coase. We describe the build-up of long term relationships which do reduce transaction costs, and we deduce the dynamical properties of networks built from our simple assumptions. In conclusion, the present model explains the metastability of employment relations in the firm, but something has to be added to it to explain the more efficient workload repartition observed in real firms.