Artificial intelligence in tax administrations: Benefits and risks of its use

Alfredo CollosaAlfredo Collosa    09 October 2020
Artificial intelligence in tax administrations: Benefits and risks of its use

The Tax Administrations (TAs) like the rest of the public and private organizations constantly seek to use new technologies to be more effective and efficient.

Within this new technology, in this comment we will refer to artificial intelligence (AI), which many TAs already use in their different functions.

Thus, for example, for information and assistance, virtual conversational assistants and chatbots are used[1], in the collection function AI is used to predict collection, in the audit AI is used in the case selection stage to detect fraud patterns and prevent breaches with risk analysis and also to carry out the audits themselves, in customs at airports with facial recognition systems, among many other uses that will surely continue to be strengthened in the future.

Therefore, the purpose of this document is to try to expose some advantages and disadvantages and disadvantages of its incorporation into the TAs, to then formulate some final reflections.


While the term "artificial intelligence"[2] It is usually associated with science fiction, in reality it is much broader and multifaceted than is believed in the collective imagination.

It was used for the first time at the Dartmouth Conference in 1956 by John McCarthy who defined it as “the science and ingenuity of making intelligent machines, especially intelligent computer programs” (Guillén Torres, 2016).

In turn, Brad Smith (2019), president of Microsoft, in his book Tools and Weapons, defines AI as a computer system that can learn from experience, distinguishing patterns within the data; patterns from which feedback is provided in order to recommend actions.

Although there is no consensus on the definition of what AI is, in general it refers to the ability of a digital system to exercise cognitive functions such as learning, interacting, creating and replicating, functions that were previously only attributed to humans.

AI can be defined as the science and engineering that makes it possible to develop machines and computer programs capable of solving problems that normally require human intelligence.

Regarding its operation[3] AI relies on automated learning to identify patterns from data, this is the engine that empowers information and works from data and algorithms that follow the following process:

1. Identify an important problem.

2. Analyze situations and prepare a study of all the variables related to the problem to be analyzed.

3. Predict the future outcome of this problem, always starting from known data and using a statistical system.

4. Once the system has all the data, provide the most feasible solution to the problem. Thus, the AI ​​learns how to solve the problem for the next similar situation it encounters.

Data analysis and AI are based on algorithms that recognize patterns through the sequential analysis of data, even from stale data or data collected long ago.

Machine learning, one of the main areas of AI, learns from the data it collected in the past and continues to collect in the present.

It's difficult to draw a clear boundary between big data and data analytics, or even between data analytics and AI, increasingly used to solve big data and data analytics problems.

At the moment, existing AI is known as soft AI, capable of performing a specific task with better performance than a human, as opposed to hard AI, which would be able to reason like a human in every respect.

Being limited in its scope, AI actually refers to a set of different techniques that allow obtaining a desired result.

Hard AI is yet to be developed and experts disagree on the possibility of developing it


AI being a disruptive and general-purpose technology, it is a tool with a power of change of great magnitude, which entails the need for a focus on responsible use, as well as the taking of necessary precautions to prevent unexpected and harmful consequences.

Responsible use of AI goes beyond not engaging in illegal practices through its use. It is about using AI in a way that does not harm minorities, that avoids human rights violations and that does not lead to an increase in the existing inequality gap, either intentionally or accidentally.

Among the consequences of the use of AI there are intrinsic risks, directly related to data, and extrinsic risks, linked to the adoption of AI in society[4].

There are at least four types of intrinsic risk to consider in the planning, programming, and implementation stages for responsible use of AI: fairness and inclusiveness, system reliability and security, user data privacy and security, and transparency and accountability. of counts

For its part, among extrinsic risks, one of the concerns has to do with the Future of Work.

The OECD estimates that, globally, 14% of jobs will be affected, while 32% will be significantly impacted. In Latin America, the IDB (2018) estimates that between 36% and 43% of jobs could be impacted by AI.

Despite common belief, these risks are not an inevitable side effect; they can be mitigated with sufficient planning, control and governance.

For this reason, leading actors on the subject such as the Organization for Economic Cooperation and Development (OECD), the Inter-American Development Bank (IDB) and other institutions speak of the responsible use of AI.

This refers to an AI development framework that aims to guarantee the ethical and transparent use of user data by adopting mechanisms that allow meeting their expectations of accountability, always in consideration of the values ​​of the organization and the laws and norms of society

On this issue the OECD[5] recommends five complementary values-based principles for responsible stewardship of trusted AI:

  • AI should benefit people and the planet by driving inclusive growth, sustainable development and well-being.
  • AI systems must be designed in a way that respects the rule of law, human rights, democratic values ​​and diversity, and must include appropriate safeguards, for example, allowing human intervention when necessary, to ensure a just and equitable society .
  • There must be transparency and responsible disclosure around AI systems to ensure that people understand AI-based results and can challenge them.
  • AI systems must function robustly and securely throughout their lifecycles, and potential risks must be continually assessed and managed.
  • Organizations and individuals who develop, implement or operate AI systems must be held accountable for their proper functioning in accordance with the above principles.

For its part, the OECD, in accordance with these values-based principles, also offers five recommendations to governments:

  • Facilitate public and private investment in research and development to stimulate innovation in trusted AI.
  • Foster accessible AI ecosystems with digital infrastructure and technologies and mechanisms for sharing data and knowledge.
  • Ensuring a policy environment that will pave the way for the implementation of reliable AI systems.
  • Empower people with AI skills and support workers for a just transition.
  • Cooperate across borders and sectors to advance the responsible stewardship of trusted AI.

In Latin America and the Caribbean the IDBalong with several leading global partners launchedfAIr LAC[6], a pioneering project that unites governments, universities and the private sector to promote an ethical use of AI.

The initiative was born with the objective of taking advantage of the immense potential of this new technology to create more efficient, fair and personalized social services for the citizens of the region, who in turn handle their data in an ethical and responsible manner.

With regard specifically to the TAs, we share what was said by Dr. Adolfo Iriarte Yanicelli[7]who affirms that in the age of knowledge, the main right that the taxpayer will have, will be the right to know how the AI ​​reaches the conclusion, perhaps controversial. Even the right to challenge the new value-added information generated.

The digitization of tax processes highlights the importance of personal data protection and privacy as we move through this digital age.

Therefore, as stated by the Tax Technology Committee[8]Taxpayers' rights must be transposed and integrated into the digital world. They say that a clear code is needed to govern the development process, the uses to which technology is put, and how the rights of taxpayers can be preserved in a digital environment.

Cristina García Herrero Blanco[9]for his part, he affirms that there are risks regarding what could be a misuse of it, which requires an adequate assessment from an ethical perspective and the adoption of a series of principles that should govern administrative action.

The aforementioned author states the principles of prudence (through which the complexity of the algorithms or the scope of the projects in which it is used is avoided, adopting the advances according to the results are safe), non-discrimination (risks that the human errors can be transferred to algorithms), proportionality, transparency and data governance to ensure its security.


The purpose of the use of AI is to transform the data into an asset of knowledge and impact tax and customs management, as well as to achieve the intelligent use of said data to transform the TAs and the way in which they will interact with taxpayers.

The combination of AI, IoT, Data Analysis and Data Analytics will give exponential benefits thanks to the collection and analysis of a large volume of taxpayer data in real time for better decision making that will positively impact various administrative areas of the AATT.

AI is great at automating repetitive tasks, increasing accuracy and efficiency, and uncovering hidden ideas and trends.

You can interpret the best way to get an answer and learn the routines that get the best result. You can automatically upload documents, understand entries, and classify them into the correct accounting codes, among other tasks.

Among its main advantages[10] stand out:

- Increase in tax collection: AI allows processing high volumes of economic information, categorizing much faster, with greater objectivity and precision than human beings, with the aim of identifying situations of non-compliance, improving control and preventing tax fraud . Revenue prediction is an area with great potential for AI application development.

- Classification of taxpayers: The AI ​​can develop a precise profile of each taxpayer based on the analysis of their past and present behavior, which allows a clearer vision of how they will conduct themselves in the future. In the context of electronic invoicing, consumption patterns of a natural person can be identified through “machine learning” technology.

- Reduction of tax evasion AI allows detecting possible irregularities through the use of algorithms in addition to performing a real-time transaction analysis to reduce fraud based on sophisticated deep learning systems.

- Diagnosis and support in decision making: AI is an ideal example of how a machine can reduce errors and speed up processes based on the use of systems capable of generating optimized strategies to solve highly complex problems and help make decisions.

- Efficiency in calculations AI has high power and efficiency in doing any kind of statistical calculation.

- Optimization of times and resources in other words, processing a large volume of information in less time.

- Support in audits: AI can be applied in audits, which allows the reduction of times since it has information in real time.

For an adequate AI function, the TAs[11] must equip themselves with human and technological assets with the capacity to:

  • Organize and provide all the data that the administration saves in a uniform format;
  • Activate the corresponding intelligence cycles, in order to transform said data into useful information for decision-making;
  • Manage feedback regarding the disseminated information, to activate subsequent processes.

The implementation of big data analysis and AI-based systems can be implemented through market products and free software packages.

Generally, the development of AI solutions involves different platforms, models and integrated algorithms depending on the type of problem and the domain (specific sector) of the solution.

Those institutions that choose to develop solutions internally must advise and strengthen their teams both in the number and type of profiles, creating for this purpose a group with experience in the management of free software, as well as capacities related to AI techniques.

This is justified by the constant and rapid transformations of the sector. With specialized external advice, the team will be able to capitalize on the progress made for the company. The model that is chosen to adopt for this advice could contemplate agreements with universities or contracts with specialized emerging companies (startups)[12].

For its part, related to how AI will affect the TAs, we share with Cristina García Herrero Blanco[13] which says that the benefits that can be achieved through the use of AI are many and in our field it should favor better tax compliance, in the sense of an easier compliance for taxpayers and at the same time fairer, in a world with fewer errors by the tax administrations.


Today, many TAs in the world use AI for multiple functions and its use is likely to continue to increase over time.

The use of AI, like the rest of the ICTs, by the TAs presents multiple benefits and opportunities for them to be more efficient and effective in their central role of improving the levels of voluntary compliance.

But you should never lose sight of the fact that Tics are a tool to obtain better results, that is, they are not an objective in itself.

On the other hand, as seen in the present, AI also implies risks for which specific regulation is required, especially for the adequate protection of the rights and guarantee of taxpayers.

At this point it is clear that AI does not act by itself, but depends on how it is "trained or programmed" by human beings, which is why they are and will be responsible for its proper functioning.

Governments must work together with the different intervening actors to guarantee the adequate use of AI, in an ethical and equitable manner, protecting the fundamental rights of citizens.

This entire process of digitization of the TAs, including the adoption of new technologies such as AI, should not be carried out in isolation, but should be integrated into the digitization of countries, within the concept of digital government.

I am convinced that digitization opens up new possibilities and opportunities for countries and their citizens to develop, and those countries that take better advantage of the digitization “train” will undoubtedly have better opportunities to develop and provide a better quality of life for their citizens.

[1] Alfredo Collosa

[2]Constanza Gómez Mont, Claudia May Del Pozo, Ana Victoria Martín del Campo. Data economy and artificial intelligence in Latin America Opportunities and risks for responsible use. This report was developed by C Minds and commissioned by the Center for Studies in Technology and Society (CETyS) of the University of San Andrés in Argentina.

[3]Artificial Intelligence Whitepaper in Tax Entities. Microsoft- PWC.

[4] Iden Note 1.



[7] Transformations and Challenges of Tax Law in the Era of Robotization and the Digital Economy: Aspects Linked to the Digital Tax Legal Relationship


[9] The use of Artificial Intelligence by Tax Administrations, a matter of principles - CIAT Blog 3/3/2020.

[10]Artificial Intelligence Whitepaper in Tax Entities. Microsoft- PWC.

[11]ICTs As a strategic tool to enhance the efficiency of the CIAT TAs pag. 209.

[12]ICTs As a strategic tool to enhance the efficiency of the CIAT TAs pag. 535

[13] The use of Artificial Intelligence by Tax Administrations, a matter of principles - CIAT Blog 3/3/2020.


Disclaimer: Content posted is for informational & knowledge sharing purposes only, and is not intended to be a substitute for professional advice related to tax, finance or accounting. The view/interpretation of the publisher is based on the available Law, guidelines and information. Each reader should take due professional care before you act after reading the contents of that article/post. No warranty whatsoever is made that any of the articles are accurate and is not intended to provide, and should not be relied on for tax or accounting advice.

You can access Law including Guidelines, Cabinet & FTA Decisions, Public Clarifications, Forms, Business Bulletins for all taxes (Vat, Excise, Customs, Corporate Tax, Transfer Pricing) for all GCC Countries in the Law Section of GCC FinTax. 

Other articles by Alfredo Collosa

like  2 Likes
views 1262

More From Articles