Statistical Assessment of the Effectiveness of Artificial Intelligence Use by organizations in Russian Regions

Purpose . Regional statistics on artificial intelligence have so far relied on the number of user organizations, leaving the outcomes of use unmeasured. The study constructs and tests a system of indexes of the effectiveness of artificial intelligence use at the regional level and identifies the factors behind its territorial differentiation . Materials and methods. The information base comprises the results of the federal statistical observation under Form No. 3-inform for 2024 and 2025 across 85 regions of the Russian Federation, including organizations’ assessments of the impact of the technology on ten types of outcomes and data on technology types, purposes of use and modes of acquisition of solutions. For each region, the proportions of positive, negative and uncertain assessments, the integral performance index as the average proportion of positive assessments, the uncertainty index and the Ryabtsev index of structural differences between the regional and all-Russian effect profiles were calculated. Descriptive statistics of variation, Spearman rank correlation, multiple regression with robust standard errors and a two-dimensional typological grouping are applied. Results . In the Russian Federation, a positive impact of artificial intelligence is most often acknowledged for the quality of business processes (62,7 percent of users) and production processes (51,1 percent), while a decrease in cost and increase in labor productivity growth are reported by only 11,5 and 24,1 percent, and from a third to half of organizations cannot assess the effect. Over the year, the shares of positive assessments increased for all results except for the decline in the number of employees, where the share fell from 15.3 to 9.1 percent. The performance index by regions averaged 29,5 percent with a coefficient of variation of 0,153, and interregional differences decreased over the year. The index is negatively related to the scale of implementation and positively to the breadth of use: an increase in the average number of usage goals per organization per unit increases the index by 6.0 percentage points when controlling for scale, implementation intensity and the proportion of uncertain estimates, the coefficient of determination of the model is 0.788. Regions with a scale and performance above the median are held by only 23.9 percent of the country's users, while 55.0 percent are concentrated in regions with a scale above the median and performance below it. Conclusion. The proposed system of indexes allows us to move from taking into account the scale of implementation to measuring its return. The effectiveness of the use of artificial intelligence is determined not by the size of the regional market, but by the depth and breadth of the use of technology within organizations, which shifts the priority of regional policy from increasing the number of users to supporting the multi-purpose development of technology.