Artificial Intelligence, Taxation, Machine Learning, Deep Learning, Big Data, Tax Administration, Big Four accounting firms, Deloitte, EY, KPMG, PwC
Reader
Abstract
Extract
of 21
Abstract
This document explores the impact of Artificial Intelligence on tax professions, its applications in tax administrations, and the concerns surrounding its use.
Extract
[...] In conclusion, in both the public and private sectors, AI has begun to transform tax practices and the professions in this field. The possibilities it offers are numerous, particularly thanks to the enormous volume of data it can process. However, to date, it is far from being able to replace professionals. It can nonetheless assist them, provided it remains under control. It also poses many challenges and its profitability seems to be correlated with the experience and volume of data it has. [...]
[...] Discrimination risks are proven. Let us not forget that AI works by following algorithms and principles determined by humans from the start. Furthermore, as long as AI is not perfectly trained, it can make mistakes. It is therefore necessary to remain vigilant and not blindly trust technology. Users must retain their right to understand the advice and decisions made about them, as well as their right to object and request a review of their case. Conducting external audits can also be an effective way to verify the correct and impartial functioning of AI.86. [...]
[...] In this sense, a third definition of AI has been proposed by a group of experts. It is a 'systemisis created by humans who, with complex goals, act in the physical or digital world by understanding their environment, interpreting structured or unstructured collected information, resonating on the knowledge derived from this information and deciding the best actions to achieve the given goal »17. Thus, currently, several approaches to the'IA coexists. Whether it is considered as a technological entity, as a science or as a system, all definitions emphasize that the finalitythe purpose of AI is to achieve tasks that could normally only be executed thanks to human intelligence. [...]
[...] In Switzerland, the digitalization of public administrations began in 2008 and AI has been progressively used53. Strengthening its use is part of the objectives defined in the "Swiss Digital Administration Strategy 2024-2027"54. The challenge is, among other things, to simplify administrative procedures, make public services more accessible and better detect frauds55. The progressive integration of AI in administrations is reflected, among other things, in the increasing importance of its use in the field of taxation. De mme, in the private sector, research and investment to develop AI applications in the field of taxation are multiplying. [...]
[...] On the other hand, an AI trained in this way also has the ability to deliver a reduced and representative sample of data from a very large number of data that were provided to it initially. This can be very useful for visualizing and analyzing data more easily and for making savings (the storage of a very large quantity of data having a cost). Finally, it is unsupervised learning that makes it possible to detect fraud or anomalies within a sample of data39. Labeldata is a very time-consuming task. Unsupervised learning therefore lightens the human workload. [...]
