CAshflow Optimization System
CAOS is a prototype simulation-based decision support system for cash-flow planning. CAOS has been designed to assist enterprises with the management of their financial negotiations and transaction and thus, balancing their corporate debt.
CAOS is a prototype simulation-based decision support system for cash-flow planning. It is designed to help users manage financial transactions with multiple vendors and clients and provide insight for further negotiations. Based on user input, the system generates and evaluates many scenarios that aim to represent possible future negotiation outcomes. An integral part of the scenario evaluation process is the UPF (Unified Planning Framework) provided by AIPlan4EU. UPF is used to model negotiation scenarios as planning problems (using PDDL) and solve them to gather information about the action sequence that will lead to the best outcome for the end user. The evaluation results are post-processed and provided as feedback to the user to aid any upcoming negotiations. Once a negotiation is finalized (or cancelled), the user can input the finalized details back to the system and continue with further evaluations.
An important role to the total reliability of the system plays the UPF platform. UPF as a unified planning framework, allowed for an easy yet precise modelling of the underlying planning problems defined within CAOS and in addition allowed the experimentation with different planning solvers. The robust approach of the framework allows for swapping between planning engines with minimal implementation effort and thus, makes CAOS able to immediately utilize upgraded versions of existing planning engines, but also to make use of new more efficient planning engines that will be supported by UPF.
A complete demonstration of the usage of the system on a realistic financial setting can be found here: Here
CAOS system reveals the great potential that such recommendation systems can have for any user who wants to manage and negotiate their financial transactions in a simple yet efficient way. Thanks to the high level of abstraction with regards to the integration of its sub-modules, the system is highly extensible and can be enhanced in the future with even more adaptive and active learning mechanisms that can allow the system to continuously improve the quality and the reliability of its suggestions.
The presented demo use-case showcases the importance of evaluating all available negotiation options in a systematic way. Successful negotiations can not only lead to the reduction of corporate debt but also to the increase of the total revenue and thus, significantly improve the financial status of end users. In addition, based on the real negotiation outcomes, users can use this information to inform the system about client behavioural changes and therefore improve the system’s accuracy.