Application of AI in Tax Administration; Global Aspect with Indian Context – Dr. Akhedan Charan

AI combines computer science and robust
datasets to enable problem-solving. AI uses computers and algorithms to mimic
human intelligence to understand, reason, and learn. To date, AI has been able
to analyze content (documents, images, sound, movies, events, etc.) to make
predictions and prescribe actions.

Artificial intelligence has gone
through many cycles of hype. However, the release of generative AI
capabilities, like OpenAI’s ChatGPT, marked a turning point. Generative AI adds
a new level of capability – Gen AI programs can generate new content and better
understand existing content. These models can learn and represent grammar,
software code, images, and many other data types. The Organization for Economic
Cooperation and Development (OECD) published a 2023 study on
AI which noted that 65% of tax administrations around the world already use AI,
40% of countries use virtual assistants, and more than half of all global tax
agencies are using AI for risk assessment.

How is Artificial
Intelligence and ML (Machine Learning) changing tax scenario?

Over the last decade of evolving regulations
in the tax landscape, Artificial Intelligence (AI) and Machine Learning
technologies have brought about an intellectual revolution. These innovations
have become indispensable tools in the competitive global market. They do not
pose a threat, but instead offer a chance to improve operations and enhance
efficiency, cost effectiveness, and customer experience.

Tax administrations globally have not ignored
this transformation. The adoption of AI and machine learning technologies is a
seismic shift reshaping how tax authorities and businesses approach tax
compliance. Businesses are realizing the power of digitization.

 Worldwide efforts to make
technological advancements in Tax Administration

Tax authorities worldwide are dealing with the
same challenges as private companies in updating old technology and automating
manual processes. It is part of the journey created by FATCA and CRS; a journey
towards greater transparency, efficiency, and accountability. Data-driven
insights guide
tax authorities and businesses towards a future where taxation fuels growth,
not burdens it.

In
recent years we have seen many countries and tax authorities quietly evaluate
and invest in AI and machine learning solutions to achieve their tax
administration goals.

FATCA (The Foreign Account Tax Compliance Act)
which was passed in 2010 requires foreign financial institutions to report
information about financial accounts held by U.S. taxpayers. Some key details
of FATCA include:

•  Requires foreign financial
  institutions to report information about financial accounts held by U.S. taxpayers
• Requires U.S. persons to
  report their non-U.S. financial assets annually to the Internal Revenue Service (IRS)
• Applies to U.S. citizens and green card
  holders living in other countries, as well as U.S.  Residents and U.S.
  corporations
• Affects non-U.S. persons who share
  accounts with U.S. persons or have U.S. persons as signatories
•  Used to detect assets, rather than income
•  Does not impose any tax, but rather
  collects and stores information for tax compliance  purposes

 

CRS (Comprehensive
Ranking System): – Common Reporting Standard (CRS): a system for the
automatic exchange of information between tax authorities

Here are some ways AI is used in tax
Administration :

1.  Detecting
fraud and Analytical Function: AI and machine learning can analyze large
amounts of data and detect patterns, anomalies and trends that may indicate
potential fraud.

The greatest transformation may be found in
focused data analysis; these technologies excel at detecting patterns,
anomalies, and trends within data. The RPA found used for workflow automation
is limited to structured data. However, AI analysis enables tax authorities and
businesses to identify potential tax evasion, make informed decisions, and
formulate data-driven policies based on unstructured data.

The analytical solutions include software examples like
Tableau, or Power BI, and using computer-assisted audit tools and techniques
(CAATTs) to improve compliance and reduce the likelihood of human oversight
errors. Rather than making decisions or conclusions based upon a limited sample
of a population, CAATTs allow for a complete review of all transactions to find
anomalies. They can also “learn” from patterns, predict which
transactions are likely to be erroneous based on past audits, and even suggest
areas of focus for human auditors.

·       
CAATTS (Compliance and Audit Automated Tracking
and Trace System) is a digital platform used by the Internal Revenue Service
(IRS) to track and manage tax compliance and audit processes. It is designed to
improve efficiency, accuracy, and transparency in the tax examination and
compliance process.

     CAATTS enables the IRS to:

     1. Track and monitor tax returns and
related documents

     2. Assign and manage audit cases

     3. Track taxpayer responses and
correspondence

     4. Store and retrieve relevant documents
and information

     5. Analyze data to identify trends and
potential compliance risks

The system
aims to enhance the overall tax compliance and audit process, reducing errors
and increasing efficiency for both the IRS and taxpayers.

1. The IRS is tackling
complex partnership tax returns in their Large Partnership Compliance program.
With the help of AI, the selection of these returns is the result of
groundbreaking collaboration among experts in data science and tax enforcement,
who have been working side-by-side to apply cutting-edge machine learning
technology to identify potential compliance risk in the areas of partnership
tax, general income tax and accounting, and international tax in a taxpayer
segment that historically has been subject to limited examination coverage.

 

2. The National Revenue Administration of
Poland is using AI driven system, called STIR, to analyze data provided daily
by banks and credit unions. This transaction analysis is being used to detect
fraud in real time rather than during annual reporting cycle.

· STIR (Secure Telephone Identity Revisited) is a
protocol designed to authenticate and verify the identity of callers in real-time,
combating spoofed or fraudulent calls. It uses digital certificates and
cryptographic techniques to ensure the accuracy of caller ID information.

        STIR enables:

        1. Authentication of caller ID

        2. Verification of caller identity

        3. Reduction of robocalls and spam
calls

        4. Improved call security and trust

 

   
    STIR is widely adopted by
telecom companies and is becoming a global standard for secure               caller authentication.

        Here are some key terms related to
STIR:

         1. SHAKEN (Signature-based Handling of
Asserted Information Using TOKENs): A framework that
 uses STIR to verify caller ID.

        2.  SIP (Session Initiation Protocol): A protocol
used for initiating and managing voice calls.

        3.  Telephone Identity Provider (TIP): A service
that issues digital certificates for caller                                authentication.

               STIR and SHAKEN have revolutionized call
authentication, making it harder for scammers to spoof
calls and providing a more secure calling experience.

               Businesses
have the ability to keep up, leveraging their own AI and machine learning strategy. In addition to the data
analysis, there are also tools that assess risk and compliance. Risk management can be aided by AI through
models which can predict potential
risks in strategies by analyzing historical audit outcomes and regulator
feedback. Regulatory compliance
monitoring leverages trained systems to monitor real-time changes in tax regulations across multiple jurisdictions
ensuring strategies remain compliant.

               Ultimately, new
technology unlocks the potential of data as a strategic asset.

2. Operations and Functions Automating
tasks: AI can automate routine tasks, freeing up human resources for more
complex tasks.

The backbone of compliance
for tax authorities and businesses in support of the front office. Implementing
new technologies such as Robotic Process Automation (RPA) with AI capabilities
removes manual procedures and creates straight through processing. Automating
tasks not only improves efficiency, but also reduces costs and increases
accuracy.

 

·       
RPA (Robotic Process Automation) is a technology
that allows organizations to automate repetitive, rule-based tasks by using
software robots to perform them. RPA tools mimic the actions of a human user by
interacting with applications and systems in the same way that a user would,
but with much greater speed and accuracy.

                RPA is
commonly used for tasks such as:

               1. Data entry and processing

               2. Customer service and support

               3. Bookkeeping and accounting

               4. Compliance and regulatory
reporting

               5. IT operations and management

               The benefits of RPA include:

               1. Increased efficiency and
productivity

               2. Improved accuracy and reduced
errors

               3. Enhanced customer experience

               4. Reduced labor costs

               5. Scalability and flexibility

               Some popular RPA tools include:

               1. UiPath

               2. Automation Anywhere

               3. Blue Prism

               4. WorkFusion

               5. Kofax

               RPA has many applications across various industries, including:

               1. Finance and banking

               2. Healthcare

               3. Insurance

               4. Manufacturing

               5. Government

 

               By automating routine tasks, RPA
enables organizations to free up resources and focus on          more
strategic and creative work.

Shortcoming
of Automization & To over come it–

 There are risks with full integration of AI or
machine learning. Without transparency and oversight, irreparable harm can fall
on the people for whom this technology is intended to help. For example, a
self-learning algorithm in the Netherlands, run by the Dutch Tax and Customs
Administration, was erroneously labeling childcare benefit claims as
fraudulent. The civil servants tasked with reviewing the cases relied on the
technology and “rubber-stamped” the erroneous flags over eight years, resulting
in thousands of families required to pay back their valid childcare tax claims.

But this Dutch example is not a large
enough red flag to stop the investments. Lessons learned from past attempts, as
well as security and ethical AI use concern have only encouraged more initiatives.
The OECD AI Principles was adopted in 2019 to promote use of AI that is
innovative and trustworthy.

Path Ahead-

The IRS continues to roll out more online functionality with the
goal to assist stakeholders. Most recently they have announced the release of a
free tool for withholding agents to validate the information being reported on
form 1042S. The tool performs a quality review of data before the
withholding agent submits to the IRS, helping ensure accuracy of information
and limiting penalty exposure in the event of audit.

       
Previously, the IRS announced that its digitization initiative has
scanned a staggering 225 times        more
forms than in the previous year. This represents a monumental leap in the
digital transformation of tax administration, aligning with the IRS’s goal of
going entirely paperless by Filing Season 2025.

               In Indian Context, the Income Tax
Department is using AI tools to assess income tax returns whose ratio of
donations to charitable trusts and political parties compared to income is
skewed. The department is reviewing prior year data which had previously not
been heavily scrutinized.

               In Australian Context Taxation Office
(ATO) has curated machine learning algorithms to consume large amounts of data
to provide tax assessments timelier. Their machines are trained to use
historical data to make improvements resulting in processes which used to take
months now taking only days. (ato.gov.au)

               Canada Sweden Vietnam context– the
Canada Revenue Agency, the General Taxation Department of Vietnam, and the
Swedish Tax Agency have also invested in AI technologies to streamline their
operational processes.

The technology keeps improving. There
are AI tools, such as Cassidy AI, which creates customized AI-assistants for
all middle and back-office functions. Feeding these tools with internal
procedures, data, and branding improves the reliability of output. This type of
augmentation will allow more tax authorities to enforce their laws while
enabling firms to remain compliant.

3. Improving
taxpayer experience: Chatbots and other AI applications can provide 24/7
support  to taxpayers, helping them navigate
the filing process and address their concerns.

Improving the taxpayer/customer experience and
service quality. To achieve these goals, AI technology is found in online UI
journeys, Q/A chatbots, document recognition, language translations of tax
laws, and better communications. Tax authorities and large businesses can
deliver quicker outcomes to benefit taxpayers/customers by automating routine
tasks.

·       
An online UI (User Interface) journey refers
to the process of interacting with a website, application, or digital product
through various touch points and stages. It encompasses the user’s experiences,
interactions, and emotions while navigating through the digital interface.

A
typical online UI journey includes:

1. Onboarding: The initial sign-up or
login process.

2. Navigation: Moving through the
website or app, finding features and information.

3. Search and discovery: Locating
specific content, products, or services.

4. Engagement: Interacting with
content, making purchases, or completing tasks.

5. Feedback and support: Receiving
help, ratings, or feedback.

6. Offboarding: The process of ending
a session or canceling an account.

 

Designing a seamless online UI journey
is crucial for user satisfaction, retention, and conversion rates. It involves
creating an intuitive, visually appealing, and responsive interface that meets
users’ needs and expectations.

·       
A chatbot is a computer program designed to
simulate conversation with human users, either via text or voice interactions.
Chatbots are typically used in messaging platforms, websites, and mobile apps
to provide customer support, answer frequently asked questions, and engage with
users.

Chatbots
can be classified into two main types:

1. Rule-based chatbots: These chat bots
use pre-defined rules and scripts to generate responses.

2. AI-powered chat bots: These chat bots
use machine learning and natural language processing (NLP) to understand and
respond to user inputs.

Chatbots offer various benefits,
including:

1. 24/7 availability

2. Improved customer experience

3. Reduced support costs

4. Increased efficiency

5. Personalization

Chat bots can be used in various
applications, such as:

1. Customer service

2. Tech support

3. E-commerce

4. Healthcare

5. Education

6. Entertainment

 

Some popular chatbot platforms
include:

1. Dialogflow

2. IBM Watson Assistant

3. Microsoft Bot Framework

4. Amazon Lex

5. ManyChat

Chatbots continue to evolve and
improve, enabling more sophisticated conversations and interactions.

However, there are risks to full
automation. Current limitations with AI supported chatbots include data privacy
and security concerns, limits on data from the last few years, and that
accuracy of responses dependent on customer prompt quality. Most problematic is
the legal risk, such as found with AI ‘Hallucination’, in which responses are
fabricated and represented as fact.

·  
AI hallucinations refer to a phenomenon where
artificial intelligence (AI) systems perceive or generate data that is not
based on actual input or reality. This can occur in various AI applications,
including:

1. Computer vision: AI may detect
objects or patterns that are not present in the image.

2. Natural Language Processing (NLP):
AI may generate text or responses that are not grounded in actual context or
facts.

3. Speech recognition: AI may
transcribe audio that is not present in the recording.

AI
hallucinations can be caused by various factors, including:

1. Overfitting or underfitting of
models

2. Biased training data

3. Adversarial attacks

4. Errors in algorithm design

AI
hallucinations can have significant consequences, such as:

1. Misdiagnosis in medical imaging

2. Inaccurate decision-making

3. Spread of misinformation

4. Erosion of trust in AI systems

To
mitigate AI hallucinations, researchers and developers are working on:

1. Improving model interpretability

2. Developing more robust training
methods

3. Implementing fact-checking and
verification processes

4. Designing AI systems that are more
transparent and explainable.

But these deficiencies are not
stopping tax authorities from investing in technology. In the US, the recently
passed Inflation Reduction Act has allowed the IRS to start their Paperless
Processing Initiative, a groundbreaking plan aimed at reducing processing times
and expediting refunds. Thanks to this initiative, the IRS has nine
taxpayer-facing voicebots in operation, in addition to 10 chatbots. To date,
taxpayers with balances due have messaged online with Collection chatbots more
than 1.6 million times.

4.  Enhancing risk assessment: AI can help identify
high-risk returns and allocate auditing resources more effectively.

5.  Streamlining operations: AI can help
digitalize and streamline workflow associated with paper returns and
correspondence.

6.  Providing personalized service: AI can help
personalize service to individual taxpayers, improving overall experience and
encouraging voluntary compliance.

In Indian context:

The
Indian government has been using artificial intelligence and machine learning
to improve tax compliance. Some of the ways AI is used in Indian tax
administration include:

               – Faceless tax assessment: AI is
used to analyze tax returns and other data to identify                                 potential fraud and errors, reducing the need
for physical assessments.

               – Automation of back-end
processes: AI and machine learning are used to automate tasks        such as data entry, scrutiny of returns, and
tax audits, reducing manual errors and                                 increasing efficiency.

               – Enhanced tax
obligation monitoring: AI is used to monitor tax obligations and send alerts to
                taxpayers, ensuring timely compliance.

               – Improved
taxpayer services: AI-powered chatbots and virtual assistants provide taxpayers
             with quick and easy access to information and
assistance.

               – Fraud detection:
AI and machine learning algorithms are used to detect and prevent fraud,      reducing
revenue loss and improving tax administration efficiency.

               – Data
analytics: AI is used to analyze large datasets to identify trends, patterns,
and insights                  that can inform tax policy and
administration.

How government strategies can address multiple areas including:

·       
Bias & Discrimination

·       
Privacy & Confidentiality

·       
Transparency & Accountability

·       
Fair Use

·       
Erosion of Trust – Hallucinations

·       
Environmental Impact – Model Size &
Resources

Sources
of This article has been used with acknowledgment and courtesy

1. Article by Sean Sutton (How is AI
& Machine Learning revolutionizing the Tax landscape?)

2. Article by Paul Dommel & Dan
Chenok (How Can AI Improve Performance in Tax Administration?)

3. WWW.ciat.org the use of artificial
intelligence by Tax Administration.

4. WWW.zeebiz.com Tax compliance trends; AI can analise extencive
data set to detect pattern that signal potential tax issues.

5. IndiaAI taxasion september 14, 2019 India said to use AI for
Faceless Tax Assessment

6. WWW.EY.com Modernizing India
corporate tax compliance; Automation, AI and Beyond. By Rahul Patni

7. Psychologyandeducation.net A study
on application of AI & Machine learning in System in Indian Taxation
System.

8. Indi AI.gov.in Artifical Intelligence in Indian taxation market
side and project growth

9. Roundtable discussion of IRS (United state internal revenue
service)

10. Discussion by IBM (International
Business Machine) center with collaboration of American University Cogod policy
center (KTPC)

11. 
Report of the organization for economic cooperation & Development
(OECD published 2023)

12. Avalara.com

Special Acknowledgment for very valuable input and assistance in
drafting to Mr. Naresh Kumar

Disclaimer: Author of this Article is working as Additional Commissioner
of State Tax in State tax Department, Rajasthan. But the fact and opinion are
not in any mean official version or authenticated. These facts have been
collected by personal efforts from various Source and opinion expressed is in
personal capacity.