The current difficulties of the National Tax Administration Agency are caused, in part, by the large influx of information and data, which is difficult to process in a relatively reasonable time, provided by the launch of several (different) programs, which consumes human resources and time, according to Ilie Resvan Dumitru, a senior adviser at the Ministry of Finance , in material published in the journal of the institution.

“In this sense, it is clear that certain problems that are currently being solved by workers can no longer be solved so effectively that underperformance is exacerbated proportionally by the infusion of low-skilled personnel, especially when the best officials leave the system (they retire, transfer or leave the job for other reasons ), not having the time necessary to pass on the accumulated experience and professional skills to the new generation,” he says.

According to this logic, says Dumitru, legal and economic sciences, as well as the sphere of their manifestation in the financial and fiscal environment (management, finance, accounting, taxation, collection, enforcement, insolvency, etc.) are carried out in practice by people with different characteristics, who must constantly to face unstructured, non-linear, dynamic and stochastic tasks.

“The performance of work duties is carried out by simultaneously taking into account both quantitative and qualitative aspects of factual situations in which elements such as intuition, experience and intelligence, or essentially ill-defined concepts, are perceived approximately as having consequences for human behavior in interaction with the economic environment,” says the adviser of the ministry.

• “Thus, the economic and social level reached by the state forces, including the National Agency of Fiscal Administration, to work intelligently to support decision-making factors, by obtaining knowledge from civil servants who are experts in specialized fields, storing them in an accepted form by a computing system and its intervention in other databases, with the help of specialized programs, to solve problems”.

• “In this way, it becomes clear why, while technology has reached such a level that it allows the development of systems that partially replace the activity of the human factor, decision-makers should begin as soon as possible to identify the problems of the institution, in order to indicate for which specific shortcomings the help of intelligent systems”.

Big data must have an AI component

What is missing or what can be completed in the process of digitization of the PNRR is the implementation of the AI ​​component in the integrated IT system represented by the Big Data platform provided for in the PNRR, he says.

“In this way, it is possible to develop and implement a tool based on a combination of hardware and software, which will be provided with the ability of self-learning based on the databases of documents/facts that are entered, interfere with both the databases of the National Agency for Fiscal Administration and with the existing algorithms of work in already implemented IT applications to identify behavioral trends based on the development of their own fiscal risk patterns,” according to Dumitru.

• “There is a need to consider the development of such an IT tool based on artificial intelligence capable of developing a powerful and effective predictive capability to correctly identify and to a large extent high fiscal risk companies involved in fraudulent commercial schemes. , money laundering, offshore tax optimization, smuggling, or who have connections with persons subject to sanctions, have a criminal past, including a history with companies written off with bad debts.

Current ANAF programs show no problems

According to this article, the shortcomings/weaknesses of the current IT applications of the National Fiscal Administration Agency, including the design of the new integrated big data IT system, are that although they pull information from an extremely full database, they are not configured or installed to the development of means necessary to solve problems is accordingly oriented to support the decisions of the leading factors by issuing

automatic appropriate decisions.

• “Existing applications do not even identify problems, but only generate indicators with inconsistencies or financial and fiscal deficiencies (fiscal data that are not correlated or have signs of fraud), indicators that should be passed through the filter and analysis of experts of civil servants. , in order to detect fiscal or other violations.”

“The difference between current IT applications and an AI-based tool is that the intelligent system is constantly self-improving, based on the typology of facts fed into the system and the patterns discovered, extracting data and information from the network necessary for the solution process problems to determine the best solutions,” he says.

• Thus, an AI-based system becomes increasingly effective at providing viable and relevant solutions the more data and patterns are fed into the system it accesses, and the more experience it accumulates from them.

• In this sense, the new Big Data system of the National Fiscal Administration should be configured on the basis of three main elements: databases, data storage and processing technologies, and intelligent data analysis.

“Data collection should not be limited to tax returns, bank accounts, tax documents and commercial registers. Data can also be collected from other sources such as IoT10 devices, social networks, bank transactions, media files, data and information stored by other institutions (SUMAL11, ANI, Courts Portal, MAI, Interpol, Europol and many others)” – he said. says .

Currently, the Ministry of Finance together with the Federal Tax Service have about 500 functional IT applications, which are gradually being developed on the platforms and technologies available at the time of their implementation.

• Their design took place gradually, simultaneously with the digitalization of the bureaucratic apparatus, depending on institutional needs and budget allocations.

“They serve the financial and fiscal processes and support processes of the two institutions, some of them are configured to detect inconsistencies in periodic reports, possible situations of state budget fraud, companies with inappropriate fiscal behavior or are configured to issue alerts on actions sensitive to fraud,” explains he.