Application of Artificial Intelligence in Tax Risk Management
Yiran Tao
Chongqing College Of Architecture And Technology, 401331, China
Keywords: Artificial Intelligence, Tax Risk Management, Application Scenarios.
Abstract: The application of artificial intelligence technology can provide more ideas and platforms for tax risk
management. In particular, the advantages of "artificial intelligence + tax risk management" in tax auditing
and screening are extremely obvious showing the inherent tax risk. This article summarizes the existing
problems in current tax risk management work based on previous work experience. The author discusses the
application of artificial intelligence in tax risk management from five aspects: upgrading management
concepts, optimizing risk management systems, innovating tax risk management methods, establishing new
artificial intelligence systems, and performing risk prediction operations.
The implementation of actual tax risk management
can use artificial intelligence technology to provide
relevant enterprises with more creative ideas. This
technology can present a great advantage in
investigating tax risks. Relevant managers can also
incorporate various tax risk influencing factors
according to the characteristics of the artificial
intelligence model. Subsequently, relevant
management personnel can also perform effective
classification and calculations to accurately obtain
tax risk content. This can not only ensure the
integration of tax risk management and artificial
intelligence, but also strengthen the risk management
and control capabilities of this type of enterprise. This
is also the essence of its changing development trend.
2.1 Principles of Artificial Intelligence
Before the application of artificial intelligence,
technicians need to clarify the technical principles
used in the technology. In general, artificial
intelligence belongs to the category of AI technology,
which has cutting-edge technological characteristics.
Its application is mainly to combine artificial and
intelligent to create different forms of computer
algorithms, and use computer technology to imitate
various behaviors of the human brain. At this stage,
the objects of artificial intelligence technology are
mainly human brain neurons. It can simulate the
internal neurons of the human brain through the
interconnection between the network and the
neurons. More importantly, the system can also
complete related tasks through network and neuron
connections. Among them, there are two most
common ways of intelligent creation. First, the whole
process of brain thinking can be fully understood and
simulated by artificial means. Second, the system can
only simulate part of the brain function, and it is only
a pure imitation. It can be created by artificial
intelligence through related function display. In
addition, artificial intelligence technology can also
identify the internal data of the system and explore
the various information hidden behind the data. When
dealing with specific problems, humans rely on a lot
of knowledge and experience, but it is difficult to
achieve good handling of hidden information that is
not easy to be discovered. To this end, staff can use
artificial intelligence system applications to
demonstrate technical advantages and technical value
(Bian, 2020).
Tao, Y.
Application of Artificial Intelligence in Tax Risk Management.
DOI: 10.5220/0011179000003440
In Proceedings of the International Conference on Big Data Economy and Digital Management (BDEDM 2022), pages 371-376
ISBN: 978-989-758-593-7
2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
Figure 1: Schematic diagram of artificial intelligence
technology principle.
2.2 Artificial Intelligence Affects Social
The development of actual information
technology d promotes the application of artificial
intelligence in many fields, and this type of
technology can show strong advanced features. The
main reason lies in the application of many new
software tools, making it easier to build complete
network applications. It can also be seen from the
application of the actual artificial intelligence system
that it can not only use multiple programming
languages, but also complete the target recognition
work. Even if it is offline, it can also train the neural
network. When the actual development and training
work is over, the system can also be transferred to the
network platform. It can be through cloud capabilities
or a PC, and more hardware and software platforms
can be integrated into it. In actual development, some
industries have ushered in more development
opportunities through artificial intelligence
applications, allowing intelligent machines to replace
manual operations. This has also led to significant
changes in China's overall labor structure, and the
impact on traditional industries is also very obvious,
and many new management models have been
extended. Currently, many corporate tax risk
management are in a state of transition. Through the
application of artificial intelligence system, the
reform work can be carried out more quickly and
effectively, and the overall risk management level of
the enterprise can be strengthened.
3.1 Traditional Tax Risk Management
In the previous implementation of the account
management system, the measures adopted for tax
management operations were "person-to-person".
With the continuous reform of the commercial
system, more taxpayers have appeared, which has
also increased the work pressure of tax
administrators. First of all, the traditional account
management model is likely to affect the
effectiveness of tax risk management. For taxpayers'
hierarchical and classified risk management, many
managers continue to use traditional inertial thinking.
This traditional method cannot clearly distinguish tax
source management matters from professional risk
management matters. Many managers continue to use
methods such as making phone calls and going to the
venue, ignoring the mining of tax-related data and
failing to accurately locate and monitor various risk
suspects. Second, tax risk management cannot run
through the entire process of tax management.
Generally speaking, tax risk management belongs to
the content of system engineering, which runs
through the work of tax collection and management.
It mainly includes pre-warning, prevention and
control during the event, and post-event evaluation. It
also includes some specific job responsibility systems
and information systems. However, it can be seen
from specific practice that risk management can
easily be used as a collection and management tool or
as a substitute for tax assessment, and cannot achieve
a good risk control effect (Yu 2020).
3.2 Tax Risk Management and
Analysis Methods Are Not
First of all, it is difficult to accurately grasp tax risk
points in the process of manual analysis. It can be seen
from the actual tax risk analysis work practice that the
risk doubts mainly come from the following aspects.
Firstly, the personal work and life experience of the
analyst. Second, the results of policy analysis. Due to
the limited sources of risk doubts, subjective judgment
does not have an effective scientific basis, and
ultimately cannot show the integrated effect of tax-
related data. Secondly, the mathematical model is
simple and the guiding role is limited. Currently, the
most commonly used tax risk analysis method is the
BDEDM 2022 - The International Conference on Big Data Economy and Digital Management
index method. The system can establish a single index
or a comprehensive index through the relationship
between the indicators, and carry out corresponding
risk scans. However, affected by artificial thinking
and the limitations of data operation, the index method
can often only construct function rules for a single risk
feature. This is also where the comprehensive index
lies, and the description logic of the risk
characteristics involved is not complicated.
Meanwhile, the selection of indicator thresholds
mainly relies on simple classification and aggregation.
It can only be applied to the data table layer, relying
too much on the subjective judgment of the modeler.
It can also be seen from this that it is very difficult to
establish comprehensive risk identification indicators
using traditional indicator methods. Finally, tax risk
management classification and classification
standards cannot be unified. In actual work
development, relevant risk analysts will generally
rank according to risk attribute information and the
resulting consequences, and then the risk task
coordinator will adjust its level. In the judgment of
specific inherent attributes, it is often based on
previous work experience, which is likely to cause risk
issues. In this way, there is no clear standardization
and standardized sorting method, and its subjectivity
is strong.
3.3 Missing Closed Loop of Tax Risk
First of all, in the implementation of actual risk
management work, it mainly relies on post-event
prevention and control. It is precisely because tax risk
management places too much emphasis on
prevention and control after the event, it is unable to
identify relevant risk content in a timely manner, let
alone monitor the whole process, which affects the
security of tax management. Secondly, the effect of
risk response feedback is not obvious enough. At this
stage, there is often no risk response link in tax risk
management. For those risks that have been
identified, it is impossible to make a comprehensive
summary and generalization. In the actual commonly
used index method, the feedback of the response
results to the index is often limited to narrowing or
widening the screening criteria, and it is impossible
to achieve model iteration. As a result, the effect of
this method on risk feedback is not obvious.
Generally speaking, the above-mentioned problems
are mainly due to the relatively short research time on
tax management risks in China, the lack of complete
regulations and procedures, and the insufficient risk
prevention and control system in China. In addition,
the tax authorities have not paid more attention to
informatization support. Although various tax
authorities have accumulated a lot of tax-related data,
the quality of the data is poor and its dispersion
characteristics are extremely obvious. In addition,
various information systems are also "independently"
in use, and it is difficult to clarify those complex risks
(Wang, 2020, Zhu, 2020).
4.1 Promote the Transformation of
Traditional Tax Risk Management
The prerequisite for artificial intelligence to play a
role is the application of big data. Only by ensuring
the full application of related algorithms can the full
training of big data be realized and the applicable
similar data can be summarized. As a result, people
can apply artificial intelligence to the field of tax risk
management in depth, and use existing empirical
analysis as the leading factor to change the traditional
tax risk management concept. People can establish a
big data tax risk management concept with data as the
core and correlation analysis as the leading factor. In
this way, “everything can be quantified can be
achieved under the influence of big data, and then
with the help of artificial intelligence, the perspective
of tax risk management can be broadened. Regardless
of the application of traditional data methods to
record various information, or the application of
unstructured data such as behavior trajectories,
artificial intelligence can summarize the connections
and laws between the data. As long as there is enough
data and abundant data connection points, the tax risk
management process and links will be more perfect.
In the actual artificial intelligence algorithm
scanning, the relevant staff should also use data as the
basis to comprehensively describe and evaluate the
risk distribution, so that classification and
classification management can be truly achieved.
4.2 Drive the Optimization of the
Traditional Tax Risk Management
First, realize the overall optimization of the tax
Application of Artificial Intelligence in Tax Risk Management
information system. With the application of modern
information technologies such as big data and
artificial intelligence in tax management, people have
put forward very high requirements on the amount of
specific tax-related data. Generally speaking, after
reaching the PB level, it can be called big data. In
consequence, if you want to truly collect tax-related
data, artificial intelligence applications are very
important. Besides, artificial intelligence also puts
forward some requirements for data sharing. It uses
specific data sources and cross-checking operations
to ensure that artificial intelligence can find hidden
value from the content of fuzzy data. This is also the
basic process of data deep mining. In general, with
the continuous application of artificial intelligence,
the construction of tax information systems will be
based on big data architecture. In this way, it can
transition from the traditional form to the new form,
and it can also achieve compatibility with various
application scenarios. In addition to the above
content, the application of artificial intelligence in tax
risk management can also ensure that the data of each
link is fully mined. In the meantime, it can also clarify
the core collection and management links, strengthen
the efficiency of collection and management, and
build a complete full-process closed-loop structure
system (Chen, 2019).
5.1 Upgrade Management Concept
At this stage, due to the relatively backward
management concepts, some companies have many
problems in tax risk management and control. The
application of artificial intelligence technology can
solve this type of dilemma, and it can truly achieve a
comprehensive upgrade of management concepts.
First of all, with the help of artificial intelligence, big
data can be regarded as the focus of management, and
the quantitative attributes of big data are used as the
basis to ensure that the tax risk management process
and indicators are more clear. In addition, the staff can
also use the network system to record personnel
information, invoice information, etc., and reflect the
specific behavior trajectory through images and
videos. Simultaneously, the application of artificial
intelligence technology can reflect the laws in internal
information and data, and then provide a basis for the
effective division of subsequent functions. Secondly,
the application of artificial intelligence can also ensure
that the thinking mode of managers is changed. The
application of traditional artificial thinking mode will
consume a lot of manpower and material resources.
But in the era of big data, managers can use audio,
video and other forms to transfer actual information to
the system platform. Managers can clarify the tax risk
points based on scientific data analysis. This kind of
thinking mode appears to be more rational and the
management efficiency brought by it is also very high.
Finally, the application of actual big data technology
can make tax risk forecasts more reasonable. It can
also reduce the transfer of multiple data to the
platform, clarify the law of risk occurrence, and
strengthen the controllability of tax risks, thereby
avoiding more economic losses for enterprises.
5.2 Optimize the Risk Management
In the implementation of actual tax risk management,
if artificial intelligence, cloud computing and other
technologies are applied, it will also place high
requirements on the technical capabilities of relevant
staff. For example, relevant staff can collect and
integrate data before applying big data. This can
provide corresponding support for artificial
intelligence system applications. In order to better
realize data sharing, people need to put forward more
requirements on artificial intelligence, and do a good
job of data inspection operations in different channels.
Only in this way can the role and value of data be truly
presented. More importantly, managers must also
make appropriate improvements to the risk
management process, and maintain the precise
attributes of artificial intelligence based on actual
conditions. This can make the data more reliable and
accurate. To achieve the above goals, relevant staff
should ensure that the management process is
transparent. For example, after tax data enters the
system platform, centralized precipitation and
conversion operations should be implemented. After
that, people need to apply and analyze again to ensure
the maximum value of the data. The application of
actual artificial intelligence in tax risk management
can apply the management method to the entire
management process to ensure that the risk is
effectively controlled (Xiang, 2019).
5.3 Innovative Tax Risk Management
First of all, technicians need to comprehensively
expand the risk analysis methods. After the artificial
BDEDM 2022 - The International Conference on Big Data Economy and Digital Management
intelligence completes the entire tax risk identification
task, the technical staff will also develop a scientific
analysis system based on the taxpayer's behavior
pattern. This ensures that both structured and
unstructured data are covered. For the enhancement of
the risk supervision process, it is also necessary to
focus on machines and staff to perform auxiliary
operations. More importantly, the staff should also
appropriately increase the frequency of use of
artificial intelligence technologies such as random
forests and decision trees based on actual conditions.
At the same time, the staff can also establish a
complete risk model, including all tax risk
characteristics and risk levels. Otherwise, in the actual
tax risk response, the artificial intelligence system can
improve the correlation characteristics of the risk
results and characteristics. Compared with the
traditional causal connection, this model needs further
improvement. Specific to a certain job, the relevant
technical personnel can establish a corresponding risk
model to achieve a comprehensive correction of the
data, and achieve a comprehensive exploration of the
risk data. What’smore, the application of this model
puts forward high requirements on technical
personnel. Enterprises should use various training
methods to strengthen the personal capabilities of
management and staff. Secondly, people have to
update the tax risk model in a timely manner. This is
mainly due to the high requirements of artificial
intelligence on the accuracy of the internal data of the
system. After the actual risk assessment operation is
over, the management personnel must also clarify the
specific improvement of the current situation of
management and control to achieve the improvement
and revision of the main model, and then perform in-
depth risk analysis.
5.4 Build a New Artificial Intelligence
All tax risk managers should clarify the role and value
of the artificial intelligence system in their daily work,
and strengthen the integrity of the ecosystem. The core
of the system is big data technology, which can ensure
the comprehensive collection of tax-related data and
establish a corresponding correction system. This can
also ensure that the management system can form a
closed-loop structure, that is, data collection,
processing and application, and so on. Meanwhile, the
application of the actual learning algorithm of the
network platform can strengthen the extensibility and
expansibility of the model, and meet various risk
requirements. In addition, the technical capabilities
that need to be applied in the construction of artificial
intelligence systems are very demanding. To this end,
the relevant management departments need to
establish a corresponding professional talent team.
This also includes risk analysts, model builders, etc.
Only by guaranteeing the improvement of technical
level and management ability can the tax risk
management procedures be more and more perfect
(Li, 2018, Li, 2018).
5.5 Perform Risk Prediction
Under normal circumstances, tax risk management
mainly comes from post-mortem analysis, which can
increase the probability of risk problems. In order to
improve this situation, people can use artificial
intelligence system design to make risk prediction
more accurate. In turn, we can avoid risks beforehand
and maintain the stable development of the company.
For example, in the development of tax risk
management of a company, relevant staff members
have established a tax inspection system based on
actual conditions. The specific procedure is shown in
Figure 2. It can learn about specific tax evasion
companies through big data applications. It can also
back up and retain the results and determine the target
company. It can also understand the company's own
tax registration. In the event of tax evasion, the system
will automatically add a corresponding score to arrive
at the final result, strengthening the accuracy of the tax
risk result. In addition, the company can establish a
key business screening system based on the tax
inspection system (Figure 3), and scientifically screen
the key businesses of the target enterprises according
to the weight of the tax bureau, which can provide
reliable financial evidence for the tax authorities.
Figure 2: Tax Inspection System.
Figure 3: Target enterprise key business screening system.
Application of Artificial Intelligence in Tax Risk Management
In summary, the development of tax risk management
is of positive significance for the development of
enterprises. In order to strengthen the scientific nature
of management concepts and methods, technicians
should understand the content of artificial
intelligence systems as much as possible. The system
is very reasonable in management ideas and
behaviors and can accurately predict specific tax
risks. Through the application of big data, all aspects
of the tax management system can be improved. This
can improve the quality of services and at the same
time make the tax risk management model develop
towards an intelligent direction.
Humanities and Social Science Research Project of
Chongqing Education Committee.
Project No. 21SKGH390.
Bian Shuaishuai. Application analysis of artificial
intelligence in tax risk management [J]. Economic
Management Abstracts, 2020(21): 51-52.
Chen Ning. Corporate tax planning to prevent corporate tax
risks and measures[J]. Taxation, 2019, 13(35): 46+49.
Li Weiren, Li Bin. Thoughts on introducing artificial
intelligence technology into tax risk analysis [J].
Taxation Research, 2018(06): 29-34.
Wang Aiqing, Zhu Kaida. Artificial intelligence technology
and tax risk management innovation [J]. Friends of
Accounting, 2020(07): 126-130.
Xiang Hongjin. Discussion on the feasibility of artificial
intelligence technology in tax risk analysis [J].
Commercial News, 2019(07): 109-110.
Yu Yiming. Research on taxation risk management based
on big data environment [J]. Modern Economic
Information, 2020(11): 123-124.
BDEDM 2022 - The International Conference on Big Data Economy and Digital Management