The work being introduced is aimed at supporting the crucial activity of deciding what is to be done, and when, within an industrial, applied, real-world situation. More specifically, matching assorted tasks to applicable production units, and deciding the priority every job is to be given. The problem, common to many different industries, arises when a considerable amount of different articles must be produced on a relatively small number of reconfigurable units. Similar issues have a strong impact on an essential concern, eminently in the textile industrial domain: satisfying the always-in-a-rush customers, while keeping accessory production costs (set-up costs, machinery cleaning costs, ...) under control, keeping at a minimum the losses related to wasteful resource-management practices, due to under pressure decision making. Given the real-world situation, where human planners tend to be the only ones considered able to tackle such a problem, the innovation hereby suggested consists of an automated, artificial intelligence based, system capable of objectively driving the search and implementation of good solutions, without being influenced by pre-existing knowledge, mimicking a powerful lateral-thinking approach, so difficult to accomplish when management pressure impedes and daunting tasks bound the human rationality. Ranking the effectiveness of a candidate solution, where pathdependency and unexpected complex effects may bias the final outcome, is not a matter trivially manageable by traditional operational research-style systems where no dynamics (recursive phenomena, feedbacks, non-linearity) appear. In order to overcome the limitations that an analytical specification of the problem imposes, the Agent-Based Modelling paradigm had to be taken into consideration. Thanks to ABM we're provided with the opportunity of in-silico experimenting every imaginable scenario, by executing the planning in a virtual lab, where the production events happen instead of simplistically being computed. In this way, we avoid following a reductionist approach, clumsily based on the usage of a static representation of the enterprise world, squashed into a cumbersome system of equations. The model has been built resorting to the Swarm toolkit (see [Bur94], [JLS99], [MBLA96]); the underlying programming language (Objective-C) made the procedure of mapping the agents involved in the process onto software objects a plain and consistent task. The problem presented belongs to the shop problems family in general, although many peculiarities make it an unconventional and distinguished one. When referring to production planning, the authors have in mind the scheduling problem rather than ERP/MRP issues. In fact, the stage of the production on which the work is focused gives the availability of raw and semi-finished materials for granted. The up- and down-streams of the supply chain are normally performed by significantly oversized equipment, in the textile industry. On the other side, core processes, spinning and weaving in particular, require peak exploitation of the available production units
