The purpose of the study is to substantiate a scientific and methodological approach to the use of artificial intelligence-based systems in marketing audit with a focus on cost minimization without deterioration of the target marketing outcome. The object of the study is the marketing audit process at enterprises, covering expenditure on data collection and preparation, advertising campaigns, customer management, content creation, pricing, promotion, distribution channels and assortment decisions. The methodology combines systems, process and cost-oriented approaches, structuring of marketing operations, author-developed scoring and formalization of a priority index for the use of artificial intelligence-based systems. The study substantiates that the economic effect of such systems is generated through the reduction of manual operations, faster detection of anomalies and duplications, lower ineffective advertising expenditure, accelerated segmentation and forecasting, fewer repeated revisions and a shorter time between problem identification and managerial response. Five types of cost are distinguished, namely direct marketing expenditure, labour-time costs, losses caused by erroneous decisions, rework costs and opportunity losses resulting from delayed response. Six marketing audit areas are structured and the role of artificial intelligence-based systems is determined for each area. An author-developed priority index integrates process automation potential, cost significance, data availability, speed of obtaining an economic effect and the risk of algorithmic error. The structural assessment assigns the highest priority to the audit of data collection, preparation and marketing reporting, with an index of 0.720, followed by the audit of advertising budget and media placement, with an index of 0.600. The scientific novelty lies in developing a cost-oriented view of marketing audit by adding a separate procedure for assessing the suitability of marketing processes for support by artificial intelligence-based systems and mandatory verification of algorithmic error risk. The practical significance of the results lies in the possibility of using the proposed mechanism at enterprises to determine the sequence of marketing process automation, revise advertising budgets, reduce manual reporting and improve the justification of managerial decisions.