BT Technology Journal  Vol 21 No 4  October 2003 59 ARMS Collaborator — intelligent agents using markets to organise resourcing in modern enterprises B Virginas, C Voudouris, G Owusu and G Anim-Ansah An efficient deployment of the BT workforce in response to the dynamic nature of customer requests for services is crucial to BT’s ability to compete in the ever-competitive telecommunications markets. The Automated Resource Management System (ARMS) is being developed to tackle this challenging problem. Its aim is to efficiently utilise the BT workforce by intelligently assigning engineers in line with forecast and real jobs. ARMS has three major functional components — Forecasting and Job Generation (FJG), Dynamic Planner (DP) and Collaborator. This paper is based on Collaborator — a computer system responsible for monitoring and supporting resource re-distribution decision-making in BT’s operational resource management units. Collaborator enhances the deployment process by allowing dynamic re-distribution of engineers among a group of customer service teams (CSTs) so that resource utilisation is improved. Collaborator focuses on balancing the workforce across multiple patches. Collaborator is formulated as a multi-agent co- ordination problem. Various software agents support the manager’s decision-making process. BT aims to further optimise resource deployment by using this novel approach, and ultimately to substantially reduce its operational costs. 1. Background and context Artificial intelligence (AI) technologies have been extensively used in the decision-making process of workforce management. The motivation for utilising such technologies in workforce management stems from the belief that automation will not just speed up the decision- making process of workforce managers, but rather produce near-optimal solutions. AI technologies offer efficient algorithms and powerful modelling tools for this purpose. For example, one could couple constraint-based modelling tools (i.e. to model complex problems) with efficient search algorithms (e.g. heuristic search methods) to solve combinatorial problems [1, 2]. The Automated Resource Management System (ARMS) is being developed to tackle the challenging problem of optimal resource distribution using AI technologies. Its aim is to efficiently utilise the BT workforce by intelligently assigning engineers in line with forecast and real jobs. ARMS has three major functional components — Forecasting and Job Generation (FJG), Dynamic Planner (DP) and Collaborator. FJG takes as its input historical job type, location, and volume records, and produces seven- day forecasts of job volumes and dummy job records (used to populate the forecast jobs). DP uses the output of FJG, coupled with the engineers’ skill profiles, availability, and their initial geographical assignments (known as their preferred working locations) to optimally assign — with little or no human intervention — area, state and skill data, thus achieving more cost-effective scheduling of the technicians. Collaborator has been built on top of the DP system and integrated with it to support multi-domain resource management functionality. While DP is responsible for resource allocation at a given level (e.g. customer service teams (CST)), Collaborator operates at a level above DP (e.g. the regional business area (RBA) level), where one RBA consists of many CST domains. While DP is concerned with the utilisation of the local workforce, Collaborator focuses on balancing the workforce across multiple patches. The aim of Collaborator is to facilitate the acquisition of additional resources to clear the workstack. This is made possible because there are surplus resources elsewhere. Collaborator enhances the deployment process by allowing dynamic redistribution of engineers among a group of CSTs so that resource utilisation is improved from a higher level perspective.