Transfer pricing at the crossroads: disruption, data, and the decade ahead

Twenty-five years ago, transfer pricing (TP) barely registered on the radar of most tax authorities, let alone most boardrooms. It was a technical discipline practised by a small community of specialists, lightly regulated, confined largely to a handful of advanced economies, and viewed by most multinationals as a documentation exercise to be completed, filed, and forgotten.

The work itself reflected that status. A benchmarking study – the centrepiece of any TP file – meant weeks of manual labour: pulling a thousand-company database shortlist, printing company documents by the ream, and working through each one manually to determine a few so-called comparable companies and establish the arm’s-length range. Six to eight weeks, minimum, for a single exercise. Painstaking, expensive, and subjective.

What has happened to TP in the intervening 25 years is one of the more remarkable stories in the history of tax – though it rarely gets told as such. It is a story of a discipline that began as a relatively contained domestic concern, evolved into a genuinely global framework, survived an extraordinary sequence of geopolitical shocks, and now finds itself on the cusp of a technological shift that will change not just how TP is done but what it is.

The formal architecture of TP begins, in practical terms, with the US. Section 482 of the Internal Revenue Code has been on the books since the 1960s, though the 1994 final regulations were what gave the principle genuine operational weight. The OECD published its first dedicated guidance in 1979 and substantially overhauled it in 1995. For most of the following decade, TP was a specialist concern of a small number of large economies while the rest of the world caught up.

India arrived in 2001 – a little late and, as anyone who experienced TP in India in those early years will attest, with considerable enthusiasm. The TP audit programme that followed was, to put it diplomatically, assertive. It produced years of contested adjustments, landmark High Court rulings, and a gradual, sometimes painful, alignment between Indian practice and the international consensus. The Safe Harbour Rules, the advance pricing agreement (APA) programme, and successive rounds of legislative refinement brought India steadily closer to the OECD framework.

Then came BEPS. The project, launched by the OECD and G20 in 2013, was the first truly coordinated attempt to stop multinationals from telling slightly different profit stories in different jurisdictions. By 2015, 15 actions had produced a new landscape: intangibles had to follow the people who actually managed them; the three-tiered documentation structure – master file, local file, country-by-country report (CbCR) – became the common language of global TP compliance, and, for the first time, tax authorities in over 50 countries were looking at the same data simultaneously. TP was not 40 different national stories. It became, broadly, one story told in 40 accents.

In 2020, COVID froze demand and disrupted supply chains within weeks. The OECD issued emergency guidance on what arm’s-length actually means when every business, everywhere, is simply trying to survive – but the guidance, however sensible, could not resolve the fundamental problem that historical comparables had become useless overnight. That same year, a price war between Saudi Arabia and Russia briefly sent US crude prices negative. For any energy group with intercompany pricing built on years of benchmarked margins, that was the week the benchmark stopped meaning anything at all.

In 2022, the war in Ukraine triggered sanctions regimes that did not even agree with each other across the US, EU, and UK – forcing multinationals to exit markets overnight and redraw supply chains around a map that had just changed.

More recently, US tariff policy has fundamentally disrupted cross-border trade and compressed margins in ways that existing intercompany pricing arrangements were not designed to absorb. Rerouted shipping through alternative trade corridors added freight costs and lead times that no historical benchmark had ever captured. Each of these disruptions forced the same fundamental question: what does ‘arm’s-length’ mean when the comparable data pre-dates a shock that changed everything?

TP survived all of it. Battered, questioned, and at times improvised – but still the framework the world reached for. That resilience tells you something about how deeply the arm’s-length principle is embedded in the international tax architecture, and how seriously the world now takes this once-quiet corner of tax.

While the disruptions were playing out, the regulatory architecture kept moving. In 2021, over 130 countries agreed to a global minimum tax. The pillar two 15% effective tax rate floor has been going live in jurisdiction after jurisdiction since 2024, and its operational weight on multinationals has proven heavier than anticipated. Companies report significant effort redesigning reporting systems, adapting ERP infrastructure, and navigating the interaction between pillar two calculations and existing TP positions. For the first time in over 40 years, the world has placed a formulary floor beneath the arm’s-length principle – not to replace it but to ensure that however a multinational prices its intercompany transactions, there is now a minimum bill.

The other half of that package, pillar one’s amount B, is still being finalised – an attempt to standardise, once and for all, how routine marketing and distribution functions get priced across jurisdictions. Chapter VII of the OECD TP Guidelines, covering intra-group services, is being rewritten to demand sharper proof that a service actually benefited the entity paying for it. Financial transactions and the DEMPE analysis of intangibles – who actually develops, enhances, maintains, protects, and exploits a valuable asset – have become the two most contested corners of the discipline.

Cutting across all of this is a shift in how tax authorities use data. Revenue authorities in India and across the globe are now routinely triangulating TP positions against customs valuations, VAT, and goods and services tax filings, as well as CbCR data. A coherent story in one place and a different story in another no longer survives contact with the data. The scrutiny is tighter, the tools available to revenue authorities are sharper, and the cost of an inconsistency – whether real or inadvertent – has never been higher.

Underneath all of this, the framework itself has not changed as much as you would think. The arm’s-length principle is still the north star; we still choose from the same toolkit of methods. What has changed is everything around that framework – the volume of transactions, the number of jurisdictions, the speed of disruption, and, increasingly, who is watching it in real time.

This is the context in which AI is beginning to reshape TP – and the most important thing to understand about that shift is that it is not happening only in corporate tax departments. It is happening simultaneously in revenue authorities, and that symmetry changes everything.

Tax administrations in multiple jurisdictions are now deploying AI agent frameworks specifically designed for TP audits. Tax authorities are not just using AI to flag cases for review – they are using it to conduct audits. Advanced systems can now automatically analyse a company’s intercompany transactions, identify the right pricing method, and find comparable companies, completing in weeks what once took a team of specialists six months or more. The practical implication is that revenue authorities can now scrutinise far more taxpayers, far more thoroughly, than they ever could before.

The corporate response to that challenge is still catching up. For most multinationals, the TP function was designed for an annual cycle: gather the data, run the benchmarks, write the documentation, file. The problem is structural – by the time documentation is filed, the transactions it defends are 12 months old, the margins may have moved, and the tax authority reviewing it may already hold more current data than the taxpayer does.

The new generation of AI-powered TP platforms is designed to change that fundamental dynamic. The more capable among them do not simply draft documentation faster or run benchmarking searches more efficiently – though they do both. What distinguishes the leading platforms is what is now being called agentic monitoring: systems that watch intercompany flows continuously, in real time, and flag the position that has drifted out of range before anyone has to ask, and maintain what is increasingly described as a living TP architecture – a defensible position that updates itself as the business, the market, and the regulatory environment evolve, rather than going stale the moment it is filed. Multi-agent workflows can now simultaneously review a transaction’s TP, VAT, withholding tax, and customs implications, identifying inconsistencies across reporting lines before they become audit findings.

For multinationals operating across multiple jurisdictions, each with its own documentation requirements and audit risk profile, the ability to maintain that kind of consistency without proportionally scaling the TP team is not an incremental improvement. It is a structural shift in how the function operates.

None of this should be read as a claim that AI is replacing TP professionals. The judgement that goes into a genuinely contested TP position – the analysis of comparability, the selection of method, the assessment of risk, the negotiation of an APA across a table from a revenue authority – remains irreducibly human. No algorithm should be sitting opposite a revenue officer, and none of them will be. The point is subtler than that.

What AI can do, and is doing with increasing reliability, is the watching: continuously, across thousands of intercompany transactions, flagging the invoice, the margin, the restructuring that does not look like its neighbours – long before a notice arrives. What it cannot do is decide what to do about it. That decision – how to weigh a precedent, how to characterise a functional shift, how to frame a position that will survive adversarial scrutiny – still belongs to professionals. The combination, when it works, is genuinely powerful: machines that see everything, professionals who know what matters.

Data governance has emerged as the critical constraint. AI systems produce only what the underlying data supports, and TP data – spread across ERPs, legal entity structures, intercompany agreements, and functional analyses – is rarely clean or consistent by default. Building the data architecture that makes AI-powered TP monitoring possible is, for most organisations, the most significant issue.

There is a broader reorientation under way. TP, for much of its history, was a compliance function. That positioning is changing. As pillar two makes the effective tax rate a boardroom metric, as ESG reporting increasingly requires companies to explain not just where tax is paid but why value is created there, and as supply chain redesigns routinely trigger TP implications that intersect with both customs and profitability, TP has acquired a seat at the strategic table it rarely occupied before. Today it sits at the intersection of tax, finance, supply chain strategy, and boardroom governance.

The CFO who once viewed TP as a year-end documentation exercise now finds it relevant to decisions about where shared service centres are located, how new intangibles are characterised from inception, how a restructured supply chain distributes margin, and what an effective tax rate disclosure actually says to an investor. That shift makes the technology investment case easier to make: a function with strategic implications deserves tools that match.

The combination of a maturing global framework, a more interventionist regulatory environment, and a new generation of AI tools that can genuinely support continuous compliance is pointing towards something the industry has been discussing in theory for years: TP as a living, always-on discipline rather than an annual exercise in retrospective justification.

The businesses that will navigate the next decade well are the ones that understood early that this discipline was no longer just about defending a price. It was about explaining where value actually lives – and having the governance, the data, and the technology to make that explanation credible, continuously, to anyone who asks.

The views expressed in this article are personal and do not constitute legal or tax advice.