When Machines Create Value: Rethinking Transfer Pricing for AI-Driven Economies

Abstract

Artificial Intelligence (AI) is transforming how multinational enterprises (MNEs) generate, manage, and capture value—posing unprecedented challenges for international transfer pricing frameworks rooted in traditional human-performed functions and tangible asset ownership. This paper explores the disruptive implications of AI-driven business models for the arm’s length principle, focusing on decentralized decision-making, data-based intangibles, and algorithmic value creation. By analyzing use cases across industries such as financial services, pharmaceuticals, and e-commerce, the study highlights core challenges including fragmented DEMPE functions, ambiguous data ownership, and comparability gaps in AI-to-AI intercompany transactions. The paper integrates extensive OECD Transfer Pricing Guidelines (2022) references, offering practical pathways for aligning AI transactions with the arm’s length standard through revised DEMPE analyses, hybrid profit-split methods, and enhanced documentation of control over risk. In doing so, it proposes a conceptual framework for recognizing economic ownership of AI-generated intangibles and advocates for OECD policy evolution to address emerging forms of digital value creation. The findings underscore the need for both tax authorities and MNEs to reimagine transfer pricing compliance in an era defined by autonomous systems, synthetic outputs, and algorithmic control.

Introduction

Artificial Intelligence (AI) is reshaping how Multinational Enterprises (MNEs) operate, transforming business models, global value chains, and intercompany transactions. AI adoption—especially Generative AI and AI Agents—is challenging traditional transfer pricing (TP) frameworks, which were designed around human-driven functions, physical assets, and conventional intangibles.

2. AI in MNE Business Models

A. Generative AI

• Creates novel content (text, images, video, code).

• Used in: Marketing, R&D, product design, drug discovery.

• Impact: Speeds up innovation and localization; raises questions about value attribution and profit sharing.

B. AI Agents

• Autonomous systems that perform multi-step tasks (e.g., booking travel, customer service, procurement).

• Used in: Financial monitoring, vendor negotiations, research.

• Impact: Function like digital employees, complicating TP due to autonomous value creation and decision-making across jurisdictions.

3. AI’s Transformation of Value Chains

AI drives efficiency, real-time responsiveness, and autonomous cross-border activity. It alters the traditional understanding of value creation, challenging existing assumptions about location, labor, and control in supply chains. This creates new complexities for TP in identifying who owns or controls AI-driven value creation.

4. Industry-Specific TP Challenges

• Finance: Centralized AI used globally for fraud detection complicates value allocation.

• Manufacturing: AI enhances Industry 4.0; value is additive but hard to measure.

• Pharma: AI accelerates R&D; IP value is spread across geographies.

• Retail: AI personalizes customer experience and optimizes inventory.

• Consulting: AI handles complex tasks traditionally done by humans.

5. Mechanisms of AI-Driven Value Creation

To align with the arm’s length principle, TP must account for:

• Cost Reduction: Savings must be fairly shared based on contribution to AI development and use.

• Profit Increase: Additional profits from AI must be allocated among contributors.

• Risk Mitigation: Value of reduced risk should be distributed to relevant parties.

• Accelerated Innovation: DEMPE functions must be applied to new intangibles created by AI.

6. Strategic Implications for TP

AI:

• Disrupts how companies forecast demand, set prices, and manage logistics.

• Is not location-bound, complicating profit attribution.

• Requires redefining economic control and ownership of algorithmic decision-making.

• Demands new TP models and valuation methods.

7. Personal Experience and Emerging Use Cases

The author shares a journey of applying AI to TP through real-world experimentation—developing AI agents to streamline analysis, documentation, and compliance. AI can:

• Draft TP reports

• Analyze vast datasets

• Automate intercompany financial analysis

• Improve efficiency, accuracy, and cost-effectiveness in TP documentation

8. AI-Agent Framework and TP Impact

AI systems consist of components such as:

• Goal Interpreter: Aligns AI objectives with company policy (like a tax manager).

• Planner & Strategist: Outlines and executes AI's actions autonomously.

• Functional Specialization: Performs domain-specific tasks (e.g., financial analysis).

• Memory & Learning: Adapts based on past actions, affecting long-term value creation.

From a TP perspective:

• Functions: AI performs diverse, simultaneous tasks.

• Assets: AI code/data are intangible and multi-use, unlike traditional assets.

• Risks: New risks arise (e.g., software bugs, non-transparency) that must be priced.

9. AI-Driven Intra-Group Transactions

Illustrated through a diagram:

• AI systems across different subsidiaries exchange services automatically (e.g., support requests, compliance reporting).

• These AI-to-AI interactions mimic shared service centers but occur without human input.

• They require TP consideration to ensure proper valuation and compliance with evolving international tax norms.

Conclusion

Artificial Intelligence (AI) presents new challenges to the traditional thinking of transfer pricing, and this paper examines the implications of AI on transfer pricing and proposes amendments to the OECD Transfer Pricing Guidelines. The guidelines should be amended to reflect value creation from AI, such as value creation from autonomous decision-making, machine-to-machine sales, joint DEMPE activities, and algorithmic risk management, because AI agents are digital workers who can perform highly valuable functions without human input. The paper illustrates how value creation from AI differs from the traditional conception of value creation, and it does not require significant human control, geographic location, or the ownership of traditional assets. AI agents are digital workers who can perform highly valuable functions without human input, thus shattering old concepts of functional analysis, risk and economic reality. This shatters old concepts of functional analysis, risk and economic reality, and therefore, the report concludes that profit splits, particularly contribution-based profit splits, are the most suitable approaches for AI. Traditional unilateral approaches such as CUP, Cost Plus, and TNMM are inadequate for AI, because they do not account for the existing practice of AI development and application within MNE groups, so the report illustrates the role of synthetic benchmarking and weighted profit splits when comparables are not available. 1) Best Approach: The report concludes that profit splits, particularly contribution-based profit splits, are the most suitable approaches for AI, and traditional unilateral approaches such as CUP, Cost Plus, and TNMM are inadequate for AI. 2) Comparability Analysis: AI is difficult to compare with other products, services or assets, because AI is constantly learning, utilising large quantities of data, and is often bespoke through algorithms, and the report illustrates the role of synthetic benchmarking and weighted profit splits when comparables are not available. Risk control, which is not people making decisions, but rather algorithms, model validation and risk mitigation, requires more documentation, such as documentation on how the software was programmed, how to turn it off, and testing, because new ownership concepts are necessary. AI creates intangibles that are difficult to own, and for example, it may be necessary to consider how data, software modifications, and value creation across borders contribute, thus the paper puts forth approaches to allocate ownership of value based on contributions at each stage of AI, rather than ownership of rights. Transfer pricing has to evolve for AI, according to the paper, so that taxation and MNEs are aligned, and also can adapt to AI in an increasingly digital and interconnected world, therefore, the paper examines the implications of AI on transfer pricing and proposes amendments to the OECD Transfer Pricing Guidelines. A. Practical Implications for MNEs: 1) Transfer Pricing Policy Development: MNEs should prioritize comprehensive AI transfer pricing policies that document the entire AI value chain, from initial development through ongoing enhancement and exploitation. This includes establishing clear protocols for valuing data contributions, algorithmic improvements, and cross-border AI services. 2) Documentation Enhancement: The research underscores the critical importance of robust documentation covering AI governance structures, risk management protocols, and functional contributions across jurisdictions. MNEs should implement enhanced record-keeping for AI decision-making processes, override mechanisms, and performance validation activities. 3) Advance Pricing Agreement Strategies: Given the uncertainty surrounding AI valuations and the limited availability of traditional comparables, MNEs should proactively engage tax authorities through bilateral or multilateral APA frameworks to establish acceptable methodologies for AI-related transactions.

References

[1] OECD. (2022). OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations 2022. OECD Publishing. [2] OECD. (2015). Aligning Transfer Pricing Outcomes with Value Creation, Actions 8–10 – 2015 Final Reports. OECD/G20 Base Erosion and Profit Shifting Project. OECD Publishing. [3] OECD. (2017). OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations 2017. OECD Publishing. [4] OECD. (2021). Transfer Pricing Guidance on Financial Transactions: Inclusive Framework on BEPS: Actions 4, 8-10. [5] OECD. (2022). Tax Challenges Arising from Digitalisation – Report on Pillar One Blueprint. OECD/G20 Inclusive Framework on BEPS. [6] OECD. (2023). Tax Challenges Arising from the Digitalisation of the Economy – Administrative Guidance on the Global Anti-Base Erosion Model Rules (Pillar Two). OECD/G20 Inclusive Framework on BEPS. OECD Publishing. [7] United Nations. (2021). Practical Manual on Transfer Pricing for Developing Countries (2021 Update). United Nations Publications. [8] United Nations. (2017). Model Double Taxation Convention between Developed and Developing Countries. United Nations Publications. [9] European Commission. (2021). Commission Notice on the revision of the guidelines on State aid for research and development and innovation. Official Journal of the European Union, C 414/01.

Copyright

Copyright © 2025 Ajit Kumar Jain. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.