Data Quality as a Key Factor in Effective Automation and Artificial Intelligence Implementation in Business Process Management

Abstract

The article presents a computational model for assessing the suitability of process data for use in artificial intelligence systems. The proposed model accounts for the nature of information defects, the specific sensitivity of algorithmic tasks, and the requirements of individual business processes. The methodological framework integrates process-based, scenario-based, comparative, and computational analytical methods. A model information dataset representing a manufacturing enterprise was constructed for practical validation. The dataset covers order processing, procurement, inventory management, production planning, customer service, and financial document workflow. Data quality is assessed in terms of completeness, accuracy, consistency, timeliness, validity, and uniqueness. The computational model incorporates a loss function, the DQAI index, task sensitivity coefficients, threshold-based assessment, and economy optimization. The analytical calculations demonstrate that identical types of data defects produce different effects across algorithmic tasks. Information incompleteness imposes greater constraints on classification operations, whereas data untimeliness proof especially critical for routing and forecasting tasks. Scenario comparison separates the direct effect of artificial intelligence implementation from the effect attributable this data preparation. Improving data quality enhances business process performance even when the algorithm configuration remains unchanged. This distinction demonstrates the limitations of applying universal assessment criteria when making automation decisions. The DQAI index accounts for the functional purpose of data and its effect on computational outcomes. The threshold model clearly distinguishes processes that are ready for digital transformation from those requiring before preparation of the information environment. The developed analytical toolkit is intended for managers, business analysts, and automation specialists responsible for selecting and justifying business processes for AI implementation. The scientific originality of the model lies in its integrated consideration of data quality assessment, algorithmic sensitivity, threshold-based admission, data cleansing costs, and the resulting business process effect.