Fragmentation of information systems prevents businesses from using AI

39% of companies use IT solutions based on artificial intelligence (AI), and their use in the financial and economic block is even lower – only 22%, according to a study by Kept. The key barrier to technology is the heterogeneity of data from information systems, the authors of the work note. At the same time, the respondents themselves (representatives of financial and economic departments of 20 companies) are not ready to single out this factor as the main reason. Kept explains: data fragmentation is the primary source and top-level factor of the obstacles that businesses talk about, as well as the complete abandonment of AI.

The heterogeneity of information system data is a key obstacle to the development of AI, according to the study “Artificial Intelligence – a Driver of Changes in Economics and Finance” by the auditing and consulting company Kept. It is based, as Kommersant was told in the organization, on the results of an initial survey and subsequent in-depth interviews with representatives of financial and economic departments of 20 industrial and trading companies. Thus, only 39% of respondents said that they use IT solutions based on AI, and only 22% of respondents use them in the financial and economic activities of companies. 17% of respondents do not plan to use technology in organizations, 22% in the financial and economic block.

“As a rule, many companies do not have a unified information landscape with uniform information bases and flows, and employees have to collect and structure a large amount of data from various sources. For AI to work effectively, it is necessary to have access to a large amount of data,” the study says. At the same time, only 4% of respondents pointed to the fragmentation of data from information systems as a barrier to the use of AI in companies themselves. Most often, respondents cited the lack of need to use AI (29%), lack of information about its capabilities (27%) and lack of infrastructure (22%). According to Pavel Zhantimirov, manager of the Kept consulting department, the conclusion about data fragmentation as a key barrier was made based on everything that the directors of economics and finance said. “The fact is that the primary source and top-level factor of the obstacles that respondents spoke about is precisely the fragmentation of data,” he explained to Kommersant.

AI can be most effectively used to solve problems located at the intersection of several blocks (finance and procurement, finance and investment, etc.) or organizational units, notes Mr. Zhantimirov, but there is no consolidated data warehouse between blocks and between organizations. According to the study, the level of AI implementation in the financial and economic block is lower than in companies as a whole, since it requires human participation in solving analytical and management problems – not all business processes are delegated to AI.

Note that the problem of data fragmentation is relevant not only for business, but also for the state. As the Accounts Chamber noted earlier, its GIS stores more than 500 thousand terabytes of data, which is difficult to use for making management decisions. The auditors pointed out the isolation, fragmentation and inconsistency of the collected data, the “patchwork” nature of the information systems and their narrow departmental or industry-specific focus (see Kommersant dated August 31, 2022).

“Data preparation is truly one of the most time-consuming processes when creating machine learning models. At the same time, large linguistic and multimodal models, due to their developed generalizing abilities, reduce the requirements for the quality of input data,” the office of Deputy Prime Minister Dmitry Chernyshenko told Kommersant. They noted that to solve the problem, the information systems of federal departments and regions are being transferred to a single cloud multi-tenant (allows to serve users from different organizations in isolation within the framework of one service) Gostekh platform, where uniform requirements for data, including formats, will be implemented.

Venera Petrova