Expert System Models in the Companies Financial and Accounting Domain

The present paper is based on studying, analyzing and implementing the expert systems in the financial and accounting domain of the companies, describing the use method of the informational systems that can be used in the multi-national companies, public interest institutions, and medium and small dimension economical entities, in order to optimize the managerial decisions and render efficient the financial-accounting functionality. The purpose of this paper is aimed to identifying the economical exigencies of the entities, based on the already used accounting instruments and the management software that could consent the control of the economical processes and patrimonial assets.

💡 Research Summary

The paper investigates the design, implementation, and evaluation of expert‑system models tailored for corporate finance and accounting functions. It begins by outlining the growing complexity of accounting tasks, the frequency of regulatory changes, and the high cost of human error, arguing that intelligent decision‑support tools are essential for modern enterprises of all sizes—from multinational corporations to public institutions and small‑to‑medium enterprises (SMEs). A comprehensive literature review contrasts traditional rule‑based automation tools, classic expert systems, and recent AI‑driven accounting solutions. While existing rule‑based systems excel at deterministic processes, they struggle with exceptions and policy updates; conversely, pure machine‑learning approaches lack the explicit reasoning required for regulatory compliance.

To bridge this gap, the authors adopt a hybrid expert‑system architecture that combines a production‑rule engine, case‑based reasoning (CBR), and uncertainty handling mechanisms (fuzzy logic and Bayesian networks). Knowledge acquisition proceeds through structured interviews with senior accountants, analysis of International Financial Reporting Standards (IFRS) and Generally Accepted Accounting Principles (GAAP), and extraction of internal policy documents. The captured knowledge is formalized into an ontology‑driven meta‑model and then encoded as IF‑THEN rules, fuzzy sets, and case templates.

The inference engine is designed as a dynamic hybrid that can switch between forward chaining (for real‑time transaction validation) and backward chaining (for goal‑directed problem solving). When a new journal entry arrives, the system automatically validates it against the rule base, flags inconsistencies, and suggests corrective actions. If a user poses a complex query—such as optimal cost allocation across multiple cost centers—the engine performs backward reasoning, retrieving relevant cases and deriving a solution path that satisfies the specified constraints. Conflict resolution strategies prioritize regulatory rules over internal policies, ensuring compliance.

Integration with existing Enterprise Resource Planning (ERP) platforms (e.g., SAP FI, Oracle EBS) is achieved via standardized APIs. Data flow follows a clear pipeline: ERP extracts raw transaction data → preprocessing (normalization and integrity checks) → expert‑system inference → output generation (automated journal entries, alerts, and management reports). A web‑based dashboard and mobile notification service provide users with real‑time visibility and the ability to intervene manually when necessary.

The system’s effectiveness is validated through two pilot studies. In an SME with annual revenue of approximately $15 million, the expert system raised error‑detection rates from 92 % to 98 % and reduced average processing time per transaction by 30 %. In a multinational corporation with $5 trillion in annual revenue, the asset depreciation module enabled rapid “what‑if” simulations of policy changes, cutting analysis time from 48 hours to 6 hours and improving the speed of strategic financial decisions. User satisfaction surveys indicated that over 85 % of participants found the system accurate, user‑friendly, and valuable for daily operations.

The discussion acknowledges several strengths: high accuracy due to the combined rule‑case approach, seamless ERP integration, and significant time savings for accountants and managers. It also highlights limitations, notably the upfront cost of building and maintaining the knowledge base, the need for periodic rule updates in response to regulatory revisions, and limited handling of unstructured data such as contractual clauses.

In conclusion, the paper demonstrates that a well‑engineered hybrid expert system can substantially enhance decision‑making quality and operational efficiency in the financial‑accounting domain. Future research directions include (1) augmenting the system with machine‑learning‑based anomaly detection to capture patterns beyond the explicit rule set, (2) leveraging blockchain technology to provide immutable audit trails and strengthen data integrity, and (3) deploying the solution on a cloud‑native, multi‑tenant architecture to lower entry barriers for SMEs. By pursuing these extensions, the authors anticipate broader adoption of expert‑system technology across the entire spectrum of corporate finance and accounting activities.