Is There Place for Artificial Intelligence in Transfer Pricing in The Post-BEPS World?

Who is paying the highest price for deviations from the arm's length principle – companies or regulators? The numbers in question seem to represent a significant burden for governments. According to estimates by the OECD, tax disputes cause annual tax revenue losses ranging from $100 billion to $240 billion. To put this figure into perspective, it is comparable to the combined GDPs of Qatar, Estonia, and Georgia.

However, these transfer pricing (TP) challenges on global economies have translated into a corporate headache too, in the form of complex tax regulations and more stringent enforcement efforts. Additionally, the increasing volume and complexity of international business transactions and sector-specific reforms have all demanded a more sophisticated approach to TP. Numerous industries, both on a global and local scale, have encountered TP challenges. To name a few: financial services, healthcare, pharmaceuticals, retail, and most notably, tech and digital.

In the financial services sector, businesses face increasing complexity in the regulatory landscape. The mandated transitions under recent IBOR reform, changing banking regulations, and the Markets in Financial Institutions Directive II have shaken things up in the industry. The healthcare sector, an already complex body of tax-exempt entities and alternative tax structures, has experienced increased dealmaking and hence transfer pricing complexities due to heightened demand for innovative solutions in light of Covid-19. Pharmaceutical companies have faced omnipresent scrutiny in transfer pricing, driven by concerns around drug pricing, tax controversies, and the goal of ensuring accessible healthcare through equitable pricing. Several prominent incidents of drug pricing and tax controversies such as the Perrigo Scandal, the Amgen Case, and 2017 Pfizer controversy which consigned Pharma to intricate compliance. Retail, in the sunlight of e-commerce's continuous uptrend, experienced the advent of a more complex transfer pricing framework due to a rise in cross-border transactions. Lastly, technology companies found themselves in the spotlight, particularly those involved in digital services, due to the intangible nature of their assets, such as patents, copyrights, and data.

Transfer Pricing

In general terms, transfer pricing refers to the pricing of cross-border, intra-firm transactions between related parties. In reality, it is a high-stakes endeavor imperative to overall strategy of MNCs – the goal? Maximizing profits, all in the realm of compliance, of course.

Picture this: one subsidiary sells goods or services to another subsidiary in a different country, determining the price at which the transaction occurs – determining profits and expenses, influencing the tax liabilities of each entity. Governments strive to ensure fair taxation, preventing profit shifting, while businesses aim to strategically allocate income to jurisdictions with more favorable tax rates.

The Base Erosion and Profit Shifting (BEPS) initiative, launched by the G20 and OECD in 2013, has created developments never before seen in transfer pricing. In 2015, final reports on the 15 Action Items of the project were released, transfer pricing documentation requirements were strengthened. MNEs were now obligated to provide comprehensive information through Country-by-Country Reporting, Master File, and Local File, enabling tax authorities to assess transfer pricing risks and identify potential profit shifting practices.

Many countries began implementing the recommendations into their domestic laws in 2016, aiming to address gaps and inconsistencies in international tax rules. BEPS initiative has also led to the revision of the OECD TP Guidelines, including economic analyses, transfer pricing methods, and arm's length principles.

Now more than ever, in tax planning, a delicate balance between compliance and optimization is more difficult to uphold when both sides of the equation are in constant competition.

Tax Planning in Latvia

Latvia, like many countries, has not been exempt from the growing scrutiny surrounding transfer pricing practices. With the introduction of a new transfer pricing legal framework under the CIT Law in 2018, Latvian companies faced additional obligations to prepare and submit transfer pricing documentation for transactions exceeding €5 million. As mentioned by the Director General of the The State Revenue Service in a recent TV interview – the regulatory body generally avoids interfering with large companies, except in matters of customs and transfer pricing. The SRS's growing interest in transfer pricing matters has led to an increase in transfer pricing audits.

Strict requirements are imposed on the format and content of transfer pricing documentation in Latvia, and failure to comply can lead to severe penalties for taxpayers. Fines can be as high as 1% of the amount involved in the controlled transaction or up to €100,000, based on the taxpayer's income or expenses for the reporting year. Penalties can be imposed for two types of violations: failure to meet submission deadlines and significant non-compliance with documentation requirements. If multiple transactions are involved, multiple fines may be applicable. It is not uncommon for cases to arise where documentation, despite containing essential information, still falls short of meeting regulatory requirements, resulting in the maximum penalty.

Additionally, a new concept called 'Consult Before' has been introduced, wherein tax authorities engage in dialogue with specific taxpayers to address concerns regarding controlled transaction pricing. If an agreement is reached during this process, no formal audit is initiated, and the taxpayer is not fined. However, failure to reach an agreement prompts the initiation of a formal audit.

These developments highlight the evolving landscape of transfer pricing in Latvia, with increased obligations, scrutiny, and the introduction of alternative approaches to address pricing concerns before resorting to formal audits.

Artificial Intelligence

AI is a branch of computer science that focuses on creating intelligent machines capable of performing tasks that typically require human intelligence. The algorithms and models that are developed enable computers to learn and reason based on data and patterns, and they themselves mimic human cognitive functions such as problem-solving, language understanding, and decision-making.

To understand how AI works, we can proceed by almost comparing it to how a chef operates: by combining ingredients (data) and cooking techniques and recipes (algorithms) that it has learned from experience, it creates a variety of dishes (in our case, predictions, recommendations, or actions).

At its core, AI revolves around the concept of machine learning, where algorithms are trained on large datasets to recognize patterns and make predictions or decisions. This training process involves feeding the AI system with labeled data, allowing it to learn from examples and make accurate predictions on new, unseen data.

Ergo the advancements in areas such as virtual assistants, autonomous vehicles, and many more – to which you may pose a very legitimate question: what’s the big deal now of AI if it has been around? We saw it transform industries from media and tech to finance and banking. The difference lies between traditional and generative AI – in the former, AI algorithms are commonly used to identify patterns within a training dataset and make predictions. The industry nowadays, however, has been seeing the advent of the latter, where algorithms are specifically designed to generate new output by learning underlying data.

Although AI solutions for transfer pricing are still in their development phase, the product itself is gaining a lot of trust in the investor landscape. Abidia, for example, is the first platform that enables multinational enterprises (MNEs) to handle transfer pricing through an AI-driven, data-centric, and automation-powered transfer pricing management system. The company successfully raised €13 million in a Series A funding round in May of 2023. While the average payout for a Series A round typically ranges from €3-8 million, Abidia's ability to secure a larger investment demonstrates the potential of the product. The market right now welcomes startups to big accounting.

Our managing partner Janis Zelmenis says: "Latvia's transfer pricing landscape is evolving, and so should our approach. AI technologies is not something we should or can afford to overlook right now, especially considering its leverage in cross-border transactions and risk mitigation. While companies may have been cautious about embracing innovation in the past, this time, it's worth keeping our doors open to welcome the benefits it can bring – we cannot afford robot-phobia."

TP-AI Integration

No industry, let alone specific business entities, is immune to the resulting increases in tax risks. However, in the scope of legal tax planning, intelligence can be tapped in a twofold manner – to both minimize exposure and create value. Rather than being a diminutive element in the equation of TP, AI takes a complementary role – with the help of automation tools, transfer pricing specialists can shift their focus from compliance-related tasks, focusing on overall supply chains and consider the strategic implications of transfer pricing decisions – which is exactly what is required in the post-BEPS world where spotlight has been shifted from technicalities towards value chains.

One way in which companies can achieve streamlined documentation and reporting is in fact through artificial intelligence - AI-powered tools can automate the preparation of TP documentation and reports. Moreover, using a web-based interface, data can be uploaded and assembled into compliant reports without duplicating efforts, also eliminating the need of cross-country customization. Apart from this, big data and robotics can assist organizations in managing the possible data overlap wherein same data set is filed and reported with other tax regulators, such as income tax authorities, customs authorities, and the registrar of companies. The primary solution for this revolves around reducing manual effort and ensuring compliance, rather than strategizing TP directly.

With the advent of big data and huge processing power, large data volumes are no longer a limitation, e.g. in benchmarking analysis, databases can integrate wider data feeds. Sampling, estimating, and extrapolating is replaced for precise and complete data sets. Real-time reporting could also become a reality in the future when tax authorities require entire data sets instead of samples. AI can continuously monitor market data, enabling organizations to stay updated on pricing dynamics. The wealth of information arising from consolidating data from diverse sources, including online marketplaces, industry reports, and financial databases, results in optimal TP decisions by considering factors such as supply and demand, market volatility, customer preferences, and competitor pricing strategies. AI powered pattern recognition also has interesting contributions for risk assessment – in conducting transactional and historical analysis companies can easily identify patterns indicative of inconsistent pricing trends or unusual profit margins.

Finding local comparables, which are generally preferred and sometimes required, can be challenging – if unavailable, foreign comparables may suffice, but these will require careful assessment of market differences between geographies. Latvia, for example, prioritizes the arm's length principle and in the case of lack of comparables, may rely on data within the Baltic States. AI can aid in assessing market differences between geographies by analyzing economic indicators, market trends, and business dynamics. It can identify and quantify variations in factors like consumer behavior, purchasing power, regulatory environment, and competitive landscape.

In the context of non-traditional comparability, organizations can leverage tech solutions such as AI powered industry clustering - instead of relying solely on geographic proximity, algorithms can cluster industries based on various parameters such as product similarity, market dynamics, and economic factors. This can help organizations

Identify comparables that may not have been apparent using traditional methods,
Address unique differences in product life cycles, intangible assets, or risk profiles,
Adapt to changes in industries, keeping analyses up to date, by capturing evolving market trends
generally expedite the whole process saving time and costs.

AI-Enhanced economic modeling can create advanced economic models that simulate the behavior incorporating relevant macroeconomic factors, market trends, and business dynamics. This tool could allow for more accurate sensitivity analysis and scenario planning - organizations would gain access to more accurate predictions regarding to transfer pricing, such as pricing trends, profitability, and potential risks that informing further decision-making.

Other solutions render themselves more interactive such as AI-Powered Semantic Search which would enable a more intelligent and nuanced understanding of a query’s context and meaning, going beyond the limitations of exact keyword matching algorithms. So - if you were to search data on EV companies in Ireland, it would show you an impressive portfolio of startups rather than mistaking them for electricians who specialize in jump-starting golf carts.

The selection of appropriate transfer pricing methods is another issue that companies cannot afford to overlook, particularly in jurisdictions where there is no prescribed hierarchy for method selection. While there are five main OECD methods for transfer pricing the use of the Comparable Uncontrolled Price (CUP) method is preferred. In the real world, numerous potential Comparable Uncontrolled Prices (CUPs) are frequently turned down due to their failure to meet essential comparability criteria like market similarities, transaction volumes, and positions within the supply chain. Even slight disparities between transactions can exert a substantial influence on pricing outcomes.

To illustrate, consider a scenario where two transactions appear identical on the surface, involving the same Product X. However, the stark contrast emerges when one transaction involves a vendor wielding market power, while the other transaction features a purchaser with the upper hand in monopolizing the market. Consequently, prices for these seemingly identical transactions can undergo a remarkable divergence, catching one off guard. In that case, by performing intelligent CUP analysis and through machine learning models - subtle differences between transactions can not only be detected but also assessed, training on historical data and incorporating relevant variables, AI models can identify patterns and correlations that may not be apparent to human analysts. Scenario analysis and what-if simulations of financial outcomes of different transfer pricing methods are thus facilitated.

Taming bias, but at what cost?

By leveraging its capabilities in data processing and analytics, a tax manager can exit the debilitating trilemma between resource allocation, incentive systems, and compliance. Not only is the trilemma robust in theory, but it also catches up quickly in the everyday of managerial bias – for instance, managers might resort to unethical practices to manipulate transfer prices for higher bonuses. When controlled for human bias with AI, the scenario is also less scary for shareholders.

However, suboptimal outcomes follow any new technology – and AI is a primary example of the phenomenon. The foremost concern of the advent of AI is data security, particularly unauthorized access to sensitive financial and business information. Governments have held more skepticism than faith in the new-found technology, what with the temporary decision of Rome to stop operating ChatGPT, the famous AI giant, in Italy over security issues.

Tech-optimistic organizations must ensure that their solutions comply with relevant regulations, both in terms of data handling and the methodologies used. Relying too heavily on technology without human oversight and intervention can quickly muddy the fixes AI is supposed to bring.

Transfer pricing solutions should take the form of value rather than mere fast-track solutions, still recognizing the importance of human expertise to assess context, and make informed judgments beyond what the AI algorithms provide. The Artificial Intelligence Act being cooked up in Brussels was passed by the European Parliament on June 14th - it focused on  concerns mostly ushering in surveillance and misinformation - although not directly relevant to tax specialists, it is indicative of future risk appetite when drawing up new regulations – as opposed to traditionally legal approaches, policymakers follow with a a “risk-based approach”.

Times are changing, and so are technologies. In a world where transfer pricing is no longer a side conversation at meetings, which seat will AI take?

Author: Maryam Sultanova

Sources: