AI-Enabled Management of Transfer Pricing Documentation: A Sustainable Governance Framework Integrating Compliance, Digitalization, and CSRD Requirements

Abstract

Tax administrations are undergoing rapid digitalisation, while sustainability requirements are increasingly embedded in corporate governance frameworks. These parallel transformations are raising new expectations for transfer pricing (TP) documentation, which must be accurate, transparent, and audit-ready. This paper investigates the extent to which artificial intelligence (AI)—specifically natural language processing (NLP), robotic process automation (RPA), and machine-learning techniques—can support a sustainability-oriented governance framework for TP documentation in multinational enterprises. Using a longitudinal case study of the OMEGA Group, operating across 21 jurisdictions, we analyse an AI-enabled documentation architecture that streamlines data extraction, enhances comparability analysis, and strengthens audit preparedness, in line with the OECD Transfer Pricing Guidelines and relevant European Union regulatory requirements. The empirical evidence indicates substantial improvements in documentation efficiency (−68.3%), a significant reduction in processing errors (−81.5%), and higher audit acceptance rates (+27%). Beyond compliance, AI-driven digital workflows contribute to sustainability objectives by reducing resource consumption, improving data traceability, and facilitating alignment with CSRD-related reporting requirements. Overall, the findings demonstrate that AI-enabled TP documentation can evolve into a strategic pillar of sustainable tax governance, provided that its outputs remain explainable, auditable, and grounded in professional judgment. The study proposes an integrated governance framework that connects digital transformation, regulatory compliance, and sustainability within contemporary TP management practices.

1. Introduction

The expansion of multinational enterprise (MNE) structures, the digital transformation of tax administrations, and the consolidation of sustainability reporting frameworks have reshaped expectations for the governance of transfer pricing (TP) documentation. Over the past decade, TP has become one of the most closely scrutinised areas of international taxation as tax authorities increasingly rely on data-driven audits, electronic reporting channels, and risk-based assessment models [1,2,3,4]. In parallel, corporate governance frameworks have begun to integrate tax transparency into broader sustainability practices, particularly under the EU Corporate Sustainability Reporting Directive (CSRD) [5,6,7]. As a result, MNEs face growing pressure to develop documentation systems that are technically robust, operationally reliable, and aligned with emerging expectations for accountability and responsible resource use. Traditional approaches—often characterised by fragmented datasets, irregular benchmarking updates, and heterogeneous local practices—struggle to meet these requirements consistently [8,9]. Recent advances in artificial intelligence (AI), notably natural language processing (NLP), robotic process automation (RPA), and machine learning, offer opportunities to standardise workflows, reduce manual inconsistencies, and improve audit preparedness.

While prior literature has examined automation in accounting, auditing, and ESG reporting, the specific contribution of AI to TP documentation remains underexplored—especially in multi-jurisdictional settings with stringent documentation requirements [3,10,11,12]. Even less attention has been paid to how AI-enabled TP processes interact with sustainability governance, despite growing recognition that tax transparency is an integral element of the governance pillar within ESG and CSRD-related reporting [10,13].

This study addresses these gaps by analysing how AI can support a sustainability-oriented governance framework for TP documentation. Drawing on an in-depth case study of the OMEGA Group (21 jurisdictions), we pursue two objectives:

(1) to examine how AI tools affect documentation efficiency, internal consistency, and audit outcomes; and

(2) to evaluate how digital TP workflows contribute to sustainability objectives through improved traceability, reduced resource consumption, and strengthened governance structures.

Building on these findings, the study proposes a conceptual model of sustainable digital TP governance. The model is grounded in empirical evidence and aligned with the OECD Transfer Pricing Guidelines and relevant EU requirements on corporate sustainability and digital accountability [1,3,14,15].

2. Conceptual Background

2.1. Transfer Pricing Documentation and Governance

Transfer pricing documentation has a dual function. From a regulatory perspective, it supports the demonstration of arm’s-length conditions and provides tax authorities with the information needed to perform risk-based assessments. From a governance perspective, documentation acts as an internal mechanism for aligning decision-making processes with OECD principles, enhancing transparency, and mitigating exposure to adjustments, penalties, and disputes [1,2].

The most recent OECD Transfer Pricing Guidelines place strong emphasis on transparency, consistency across jurisdictions, and sound comparability analysis as core elements of TP governance [1,2]. As compliance expectations intensify, documentation becomes not only a technical requirement but also a key instrument for demonstrating disciplined and coherent application of TP policies throughout multinational groups.

2.2. Digitalization and AI in Tax and Compliance Workflows

Digitalization has considerably changed the architecture of tax compliance. Tax administrations make growing use of analytics, pre-populated reporting, and automated risk-assessment tools [16,17,18].

In response, MNEs increasingly deploy integrated digital workflows to handle large volumes of structured and unstructured data and to standardise documentation practices.

AI technologies—NLP, RPA, and machine learning—can automate data extraction, highlight inconsistencies, and generate structured documentation drafts compatible with OECD-compliant templates [19,20]. These tools enhance data integrity and reduce the time required to assemble functional analyses, benchmarking studies, and reconciliations.

However, the deployment of AI in this field raises specific requirements regarding explainability and auditability, particularly in jurisdictions where documentation is examined in detail, and TP positions can be challenged in administrative or judicial fora. Recent developments in responsible-AI standards, including ISO/IEC 42001 and the broader literature on explainable AI, underline the importance of making AI outputs interpretable, traceable, and defensible [21,22].

2.3. Sustainability Governance, CSRD, and Tax Transparency

CSRD broadens the scope of governance and transparency requirements applicable to large EU-based entities. Under this framework, tax governance is treated not only as a compliance matter but also as a component of responsible corporate conduct, contributing to accountability, ethical behaviour, and efficient resource use [23,24,25].

TP documentation interacts with sustainability objectives through several channels:

• Reducing the use of physical materials (paper, courier transport);
• Reinforcing traceability and internal control;
• Promoting cross-border consistency in qualitative and quantitative disclosures;
• Supporting the transparency of tax policies within ESG governance architectures [1].

These developments are consistent with a number of Sustainable Development Goals (SDGs), notably SDG 9 (industry, innovation and infrastructure), SDG 12 (responsible consumption and production), and SDG 16 (peace, justice and strong institutions) [24]. Within this setting, digital TP documentation acquires a dual role: it serves compliance and, at the same time, contributes to sustainable governance structures at enterprise level.

3. Materials and Methods

3.1. Research Philosophy and Rationale

The study is anchored in a pragmatic research philosophy, which is particularly appropriate for contexts where technological innovation intersects with regulatory obligations and sustainability-oriented governance [6]. A pragmatic stance permits the combined use of qualitative and quantitative evidence to analyse how AI-enabled TP documentation functions both as a compliance instrument and as a mechanism influencing transparency, accountability, and the sustainability profile of corporate governance.

This perspective is in line with an emerging body of work that recognises AI in tax compliance as a socio-technical phenomenon: documentation processes are shaped not only by algorithms and data structures, but also by organisational routines, managerial judgement, and the interpretation of regulatory frameworks [5,18,22]. A single methodological lens would therefore be insufficient to capture the interdependence between the technical functionality of AI tools and the institutional environment in which they operate. Methodological pluralism offers a more adequate picture of how digital TP documentation is produced, reviewed, governed, and evaluated by internal and external stakeholders [6].

The design is also dictated by the research question. Understanding how AI modifies documentation quality, efficiency, and governance integrity requires evidence on both operational performance (cycle times, error rates, alignment of benchmarking analyses) and organisational behaviour (adoption patterns, interpretative practices, perceived risks, sustainability implications). A pragmatic mixed-methods framework enables the integration of these dimensions.

3.2. Mixed-Methods Architecture

To capture the multi-layered effects of AI-enabled documentation, we adopt a convergent mixed-methods design in which structured quantitative measurements are combined with qualitative insights from interviews and documentary analysis [6]. The two strands were developed in parallel and brought together at the interpretation stage to support triangulation and analytical coherence.

(1) Quantitative strand

The quantitative component focuses on operational and governance-related indicators before and after AI implementation, such as:

• Documentation cycle time across 21 jurisdictions;
• Numerical consistency between ERP data, benchmarking analyses, and final TP positions;
• Adjustments computed using arm’s-length interquartile ranges [1,2];
• Audit acceptance rates in jurisdictions with a high degree of formalism [18,21];
• AI explainability metrics (XAI layers) used for traceability and auditability [22,23];
• Sustainability indicators including reductions in paper consumption and CO₂ emissions linked to printing and couriering [24,26].

These indicators were derived from system logs, internal reports, benchmarking refresh cycles, and anonymised audit outcomes, with the aim of capturing year-on-year improvements in both compliance quality and sustainability performance.

(2) Qualitative strand

The qualitative strand comprises:

• Semi-structured interviews with tax managers, TP specialists, compliance officers, and IT staff [6];
• Analysis of TP files, Local and Master File versions, intercompany agreements, management representations, and ERP extracts [1,2];
• Expert assessments of AI-generated narratives and tables, indicating segments requiring human validation or correction [22,23].

This strand explores how employees understand and use the system, how human–machine interaction is negotiated in practice, how documentation governance evolves, and how sustainability concerns are integrated into routine compliance work.

Integration

The two strands were integrated by systematically cross-checking quantitative and qualitative findings. Reductions in error rates, for example, were examined alongside interview narratives describing improved version control and less manual reconciliation. Similarly, environmental indicators were compared with qualitative reflections regarding reduced printing and shorter audit cycles [24,26]. In this way, both system-level performance and human experience are reflected in the results [6].

3.3. Case Study Methodology

Empirical analysis is based on a single in-depth case study of the OMEGA Group, selected via purposive sampling. OMEGA is an appropriate case for at least three reasons.

First, the group displays advanced digitalization maturity, having implemented a multi-country ERP ecosystem that supports AI ingestion and data structuring. This infrastructure allows us to examine AI performance in a context where technological frictions are relatively limited [20,21].

Second, OMEGA has substantial TP exposure, operating in 21 jurisdictions, several of which are characterised by strict, formalistic documentation requirements (notably Romania and Poland) [18,20]. This provides a rich empirical basis for analysing audit outcomes, compliance risks, and cross-border documentation consistency.

Third, the group belongs to the early adopters of AI-enabled TP documentation tools, which makes it possible to compare pre- and post-AI documentation cycles over multiple fiscal years. This longitudinal dimension is essential for capturing both sustainability effects and governance transformations [1,26].

The case study follows Yin’s methodological principles, ensuring reliability and validity through:

• Systematic coding of qualitative materials [6];
• Explicit links between evidence and analytical propositions [6];
• A documented chain of evidence;
• Use of multiple data sources to support replication logic [6].

This design enhances analytical generalisability to other multinationals operating in complex regulatory settings.

3.4. Data Sources

Data collection proceeded in several stages and resulted in three main categories of evidence.

(A)

Documentary evidence

Over 200 TP-related documents were reviewed, including:

• Intercompany agreements (service, royalty, management fee, cost allocations);
• Audit reports and correspondence with tax authorities;
• ESG/CSRD governance documents [10,16,26];
• Internal Standard Operating Procedures (SOPs).

All documents were anonymised and indexed through a metadata matrix capturing consistency, version control, document lineage, and traceability, which allowed for systematic comparisons between baseline (manual) and AI-assisted documentation.

(B)

System-level data

System evidence comprised:

• AI-generated documentation drafts;
• NLP extraction logs;
• XAI explainability layers associated with individual data points [22,23];
• Benchmarking refresh cycles [3,14,18];
• Monthly “compliance health reports” produced by the system;
• Time logs, error logs, and pre-validation metrics.

These data support a granular assessment of efficiency, accuracy, and compliance gains.

(C)

Interview data

Six semi-structured interviews (45–90 min) were conducted with:

• The Group Tax Director;
• The Chief Information Officer;
• Two transfer pricing managers;
• Two local compliance officers.

The interviews covered perceptions of AI reliability, changes in workflow coordination, patterns of organisational resistance or acceptance, audit defensibility, and alignment with sustainability goals. Transcripts were anonymised, thematically coded, and cross-referenced with documentary and quantitative material [6].

3.5. Analytical Strategy

Each dataset was analysed with methods appropriate to its characteristics. Quantitative indicators were examined using descriptive statistics and time-series comparisons. Qualitative transcripts were coded using a hybrid deductive–inductive approach inspired by Braun and Clarke’s thematic analysis [2]. System logs were integrated with documentary evidence to reconstruct the end-to-end documentation workflow. This multi-layered analytical approach provided the backbone for triangulation and enhanced the robustness of the conclusions [6].

3.6. Validity, Reliability, and Limitations

Several measures were taken to strengthen the credibility of the study. Construct validity was supported by reliance on multiple independent data sources (documents, interviews, system logs) and by aligning coding procedures with OECD, EU, and national TP frameworks [10,17,18]. Internal validity benefited from longitudinal comparison and member-checking with interview participants [26]. External validity is reinforced by the structural similarity between OMEGA and many multinationals operating in Central and Eastern Europe [20]. Reliability was addressed through detailed documentation of coding procedures, audit trails, and system-generated logs [6].

Limitations remain. Access to highly confidential materials was necessarily constrained; digital maturity varies across subsidiaries; and AI governance regulations are evolving [11,14]. These aspects do not undermine the analytical value of the study but indicate areas for further research.

4. Results

4.1. Baseline Documentation Challenges

Before the implementation of the AI-enabled documentation system, the OMEGA Group operated within a transfer pricing environment characterized by fragmentation, heterogeneity, and increasing pressure from tax administrations to provide highly structured, evidence-based, and internally consistent TP documentation [18,19]. Although OMEGA is a technologically advanced multinational, its transfer pricing (TP) compliance processes had evolved organically over time, leaving a legacy of manual practices embedded in local routines. Understanding this baseline is critical because the quality, structure, and internal governance of TP documentation fundamentally determine the effectiveness of an AI-enabled system once introduced.

Fragmentation of data and processes

The first and most visible challenge was the fragmentation of data sources. Key elements of the Local Files—functional analysis, financial data, cost allocations, intercompany service descriptions, and benchmarking—were stored across various systems that did not communicate with one another. Finance used SAP ERP modules, legal teams used document-sharing platforms, TP staff stored benchmarking studies in local drives, and operational teams held SOPs and managerial reports outside any structured repository.

This fragmentation meant that documentation drafting started with an extensive “data hunt.” Interviewees repeatedly described the process using phrases such as “searching for data that should already have been centralized” or “trying to combine five datasets that did not speak the same language.” In audited jurisdictions such as Romania and Poland, missing or inconsistent data created immediate risks: incomplete benefit tests, outdated functional descriptions, or unexplained variations in profitability could easily trigger audit adjustments [19,21].

Manual compilation of local files

The second baseline weakness involved the manual preparation of Local Files, which often required between 10 and 20 hours of work per jurisdiction. The process relied on copying and pasting large text blocks from prior-year reports, editing sections inconsistently, and manually formatting tables. This approach introduced substantial human error.

In the empirical dataset, OMEGA identified:

• Numerical discrepancies between tables within the same Local File;
• Functional descriptions that did not match ERP-recorded activities;
• Benchmarking analyses that were reused without annual refresh [3,4,18];
• Internal inconsistencies across jurisdictions performing similar functions.

In Romania, for example, the functional analysis described the subsidiary as a limited-risk distributor, while ERP records and management reports indicated that local teams occasionally engaged in minor marketing and customer relationship management activities. Such discrepancies weakened the defensibility of the Local File because tax authorities often rely on functional characterization to evaluate the appropriateness of the transfer pricing method [18,21].

Narrative–numeric misalignment

Another recurrent problem was the misalignment between narrative sections and numerical data. Because many reports were drafted manually and updated incrementally year after year, numerical values in tables often reflected current-year financials, while narratives referenced prior-year events or outdated processes.

This situation was especially common in sections describing:

• The business restructuring history of the group;
• The rationale for intercompany service fees;
• Descriptions of intangible asset ownership;
• Local operational capabilities.

In one audited year, OMEGA’s Romanian Local File contained a mismatch between the narrative describing a centralized IT service model and the actual cost allocation keys used in ERP, which applied a headcount-based apportionment. During the audit, authorities requested clarification to reconcile the narrative with the actual cost structures—an operational risk directly linked to the lack of version control and automated narrative alignment [22].

Benchmarking gaps and outdated comparables

Benchmarking studies, a crucial component of TP documentation, suffered from annual refresh inconsistencies. OMEGA’s baseline review revealed that 30–40% of benchmark studies used for certain service categories had not been refreshed in the year under examination. This misalignment increased the likelihood of margins falling outside interquartile ranges [18].

The risk was particularly significant in industries with volatile financial cycles—automotive components, electronics, and logistics, all relevant to OMEGA’s business. Furthermore, comparables often included companies that were no longer functionally aligned with the tested party, either due to changes in their business models or the availability of updated datasets. Without automation, these issues were difficult to track [3,4].

Audit exposure and formalistic documentation requirements

In several jurisdictions, especially in Central and Eastern Europe, tax authorities follow highly formalistic documentation requirements [19,21]. Romania, Poland, Hungary, and Greece expect Local Files to demonstrate:

• Strong narrative coherence;
• Economic substance behind service fees;
• Annual comparability updates [18];
• Consistency between financial results and the selected TP method;
• Robust evidence for benefit demonstration.

The empirical record from OMEGA shows that before AI implementation, Romanian authorities requested clarifications in three consecutive years due to inconsistencies in functional analysis, missing references to specific cost drivers, and incomplete evidence for management services [19,21]. These weaknesses created both operational pressure and the perception among TP teams that audit cycles were becoming increasingly unpredictable and time-consuming.

Cross-jurisdictional inconsistencies

Perhaps the most systemic baseline challenge stemmed from inconsistencies across jurisdictions. Because each subsidiary developed its own documentation practices, the multinational lacked a unified template. The same service—the management support service—was described using different wording, different subsections, and different structures for benefit demonstration across countries.

These inconsistencies undermined both audit defensibility and internal governance. Senior management could not easily review TP compliance status globally because documentation did not follow a common analytical framework [18,19].

Governance weaknesses linked to manual processes

Finally, the baseline environment was characterized by weak internal governance due to limited transparency. Without standardized templates, versioning protocols, and traceable audit trails, it was difficult to ensure accountability or to monitor the completeness and consistency of Local Files.

Interview evidence indicates that reviews often consisted of “scanning for major red flags” rather than performing structured quality assessments. As one TP manager noted, “the risk is not the calculation—the risk is the documentation around the calculation” [22].

Consolidated interpretation of baseline challenges

The pre-AI documentation environment at OMEGA was thus characterized by:

• Excessive manual labor;
• Inconsistent documentation structures;
• Fragmented data;
• Minimal governance controls;
• Insufficient sustainability integration [24,26];
• Variable audit robustness [19,21].

These weaknesses form the baseline against which AI-enabled improvements must be interpreted, highlighting the necessity and value of transitioning to a structured and automated TP documentation framework.

4.2. Workflow Transformation After AI Implementation

The introduction of an AI-enabled documentation system within OMEGA fundamentally reshaped the architecture of transfer pricing (TP) compliance and the internal logic through which information is gathered, validated, and communicated to tax authorities. Unlike minor process improvements or incremental digital tools, the transition to an integrated AI environment produced a structural break in the way the organisation interprets and operationalises OECD standards [18], manages cross-border consistency, and consolidates its governance responsibilities. The empirical evidence collected for this study illustrates a shift from a historically fragmented, labour-intensive workflow toward a coherent documentation ecosystem that reflects both regulatory expectations and contemporary sustainability requirements.

4.2.1. Reconfiguration of Data Collection and Validation Mechanisms

Prior to automation, OMEGA relied on an extensive chain of manual tasks to assemble TP documentation. Local teams extracted financial data from ERP modules, searched for relevant intercompany agreements, reconciled functional information with operational reports, and often struggled to locate the most up-to-date benchmarking studies. This “search-and-compile” pattern resonates with the wider challenges described in the literature on digital tax administration, where data fragmentation is consistently identified as a key obstacle to quality compliance [19,20].

The deployment of NLP-based extraction modules within OMEGA shifted this paradigm. The system continuously processes contract repositories, Master File versions, and ERP outputs, allocating each data point to a predefined structure. As a result, inconsistencies that previously required days of manual review can now be identified immediately. Because the classification rules directly reflect the OECD Transfer Pricing Guidelines [18], the quality of extracted information aligns more closely with the requirements used by tax auditors when evaluating TP positions.

This transition from episodic, human-led data collection to continuous, automated data ingestion not only reduced operational pressure on TP staff but also fundamentally improved the organisation’s ability to maintain coherent evidence trails across fiscal years.

4.2.2. Standardisation of Narrative Structures and the Role of Linguistic Models

The second major transformation concerns narrative standardisation. In the pre-AI environment, OMEGA’s Local Files varied significantly by jurisdiction, especially in the articulation of functions, risks, and intercompany services. These inconsistencies arose partly from differences in drafting practices and partly from the absence of a unified conceptual template. Prior research highlights that inconsistent narratives limit the reliability of documentation and exacerbate audit risks, particularly in jurisdictions that place strong emphasis on the internal coherence of TP reports [18,19].

The AI system introduced at OMEGA uses a specialised language model calibrated on legal-economic corpora, similar to approaches documented by Chalkidis et al. [5]. This model is designed to generate structured, internally consistent narratives aligned with OECD requirements and with the factual profile extracted from corporate systems. Importantly, the system does not replace human judgment; instead, it produces a first draft that adheres to a logical, tax-oriented structure, which TP specialists subsequently refine.

Interview evidence indicates that users perceived this structured starting point as a significant improvement. The process eliminated the redundant redrafting of annual narratives and reduced interpretative divergences across subsidiaries. In effect, AI did not standardise content mechanically but regularised the logic of exposition, providing a stable platform for analytical work.

4.2.3. Automated Refresh of Benchmark Analyses

Benchmark studies represent one of the most sensitive components of TP documentation, given their central role in substantiating the arm’s-length nature of intercompany transactions. As noted in global industry surveys [21], outdated comparable sets remain one of the most frequent sources of adjustments during audit. OMEGA’s baseline situation reflected the same challenge: in several instances, benchmark updates were postponed due to lack of resources or delayed availability of public financial statements.

The AI system addressed this vulnerability by integrating directly with commercial databases such as Amadeus and Orbis [3,4]. Each time new financial data becomes available, the system evaluates the comparables used in previous studies and signals whether they remain functionally aligned, financially relevant, and statistically robust. Where necessary, the system recommends updates and recalculates interquartile ranges used for economic testing [18].

This automation provides a double advantage. First, it ensures that benchmark sets remain up-to-date without requiring extensive manual work. Second, it strengthens audit defensibility, particularly in jurisdictions where authorities increasingly question the currency and relevance of comparables, such as Romania, Poland, and Italy. The approach is fully consistent with OECD’s emphasis on maintaining current and reliable financial data in comparability analyses [18].

4.2.4. Cross-Border Consistency and Reduction in Inter-Jurisdictional Divergence

A recurring difficulty identified in the empirical study concerns discrepancies between Local Files prepared in different countries. Although subsidiaries performed similar operational roles—frequently that of limited-risk distributors—their documentation described these roles using different terminologies, varying levels of detail, and divergent interpretations of group-wide policies. Such inconsistencies create room for misinterpretation and may trigger questions relating to profit allocation or the alignment between functions and remuneration [18,19].

The introduction of AI fundamentally mitigated this problem. Because the system relies on a unified conceptual taxonomy of functions, risks, cost drivers, and service categories, the resulting documentation displays a level of structural coherence that was previously unattainable. Even when local teams contribute jurisdiction-specific insights, the overarching logic remains consistent. The system effectively acts as a semantic stabiliser, ensuring that similar activities are described in comparable analytical terms across all jurisdictions.

Users emphasised that this cross-border alignment substantially facilitated internal reviews and strengthened the organisation’s position during audits, where authorities increasingly conduct cross-jurisdictional comparisons of TP documentation.

4.2.5. Governance Enhancements Through Compliance Health-Check Dashboards

The automation of documentation is closely tied to improvements in internal governance. OMEGA adopted a set of compliance dashboards integrated directly into the AI environment. These dashboards provide real-time visibility over:

• Missing or outdated agreements;
• Gaps in Master File–Local File alignment;
• Expired or incomplete benchmark studies;
• Inconsistencies between ERP results and profitability indicators;
• Countries requiring immediate remediation.

This development aligns with broader trends in corporate governance and ESG reporting, where transparency, traceability, and longitudinal monitoring of compliance processes are increasingly important [10,16,26]. For the first time, senior management could observe the documentation status of all subsidiaries in a structured and comparative format, rather than relying on heterogeneous manual reports.

By incorporating auditability principles drawn from ISO/IEC 42001 [14], the system enhanced OMEGA’s capacity to justify internal processes, respond to queries from tax administrations, and maintain consistency with CSRD reporting obligations [10].

4.2.6. Interaction Between Digital Transformation and Sustainability Outcomes

The empirical analysis highlights a less discussed but significant dimension: the positive environmental and sustainability effects generated by digital TP documentation. Eliminating the reliance on printed documentation and reducing cross-border shipping of audit packages contributed to measurable resource savings. OMEGA reported a reduction of approximately 78 kg of paper annually and the avoidance of roughly 2.3 tons of CO₂ previously associated with printing and courier services.

These results are in direct alignment with the sustainability objectives established under Agenda 2030 [24] and the accountability principles promoted by CSRD [10]. In this respect, AI serves not merely as a compliance tool but as a vehicle through which the fiscal function contributes to the broader sustainability strategy of the multinational—an interaction that the literature has only recently begun to articulate [23].

4.2.7. Integration of AI Governance and Explainability Requirements

An important dimension of the transformation concerns explainability. Contrary to perceptions that AI systems operate as opaque decision engines, OMEGA implemented an XAI framework that allows users to inspect the justification behind each generated paragraph, financial mapping, or benchmarking conclusion.

This requirement reflects contemporary debates on responsible AI, where explainability is considered essential for maintaining trust and ensuring regulatory compliance [22,23]. By offering these transparency layers, the system avoids the black-box concern often associated with machine learning in regulated environments and aligns with the governance expectations that will become binding under the EU AI Act [11].

4.2.8. Consolidated Interpretation

Overall, the transformation observed at OMEGA demonstrates that AI deployment in transfer pricing is most effective when integrated into a broader system of governance, benchmarking integrity, data quality, and sustainability alignment. The empirical evidence shows consistent improvements in internal coherence, auditability, operational efficiency, and environmental performance.

These developments support the view that digitalisation in the fiscal function can generate not only compliance benefits but also governance and sustainability enhancements, echoing positions increasingly reflected in the professional and academic literature [19,26].

4.3. Quantitative Performance Improvements

The quantitative findings emerging from the OMEGA case study illustrate the scale and depth of the transformation triggered by AI-enabled TP documentation. While qualitative accounts help illuminate perceptions and workflow dynamics, a rigorous quantitative assessment provides a clearer view of the magnitude of operational, financial, governance, and sustainability benefits attributable to the new system. This section synthesises system-generated evidence, ERP-derived indicators, longitudinal performance data, and audit outcomes, placing them in dialogue with international benchmarks reported in global surveys [21] and the expectations articulated in the OECD guidelines [18].

4.3.1. Reduction in Documentation Cycle Time

One of the most pronounced quantitative effects concerns the reduction in documentation cycle time. Before implementation, the preparation of a Local File required, on average, 14.6 hours per jurisdiction. This duration reflected not only the inherent complexity of assembling TP documentation but also inefficiencies associated with repetitive manual tasks—data extraction, narrative alignment, cross-checking of financial results, and benchmarking update cycles.

Following AI integration, the average cycle time declined to 4.6 hours, marking a reduction of 68.3%. The decrease is significant not merely in statistical terms but also in its organisational implications. At group level, across the 21 OMEGA jurisdictions, the cumulative time savings per year exceed 210 hours of specialised TP labour—equivalent to several weeks of full-time analytical capacity that can be redeployed to higher-value activities.

The mechanisms behind these improvements can be traced directly to the workflow innovations described earlier. Automated data extraction eliminated the need for repetitive searches; pre-structured narratives reduced drafting time; and the real-time alignment between benchmarking tables and ERP data removed the iterative correction cycles that previously consumed substantial resources. These results are consistent with the literature on automation in accounting and audit functions, which reports time reductions between 40% and 70% for routine compliance tasks [5].

4.3.2. Decrease in Error Rates and Internal Inconsistencies

The second major quantitative improvement relates to documentation accuracy. Before the transition to AI, the error logs maintained by OMEGA indicated a recurring pattern of inconsistencies between narrative descriptions, intercompany service lists, and financial outcomes. These inconsistencies required multiple rounds of internal review and represented a non-negligible audit risk, particularly in formalistic jurisdictions.

Post-implementation, error rates decreased by 81.5%, representing the most substantial qualitative-to-quantitative shift in the dataset. The system’s validation layer identifies numerical mismatches, missing references, outdated profitability ranges, and divergences between functional descriptions and ERP-derived operational profiles. Because these alerts are issued at the drafting stage, inconsistencies rarely reach the final documentation.

This improvement is consistent with emerging scholarship emphasising the role of AI in enhancing data integrity and reducing compliance-related operational risks [22,23]. It also aligns with the OECD’s emphasis on internal coherence as a key indicator of TP documentation reliability [18].

4.3.3. Audit Acceptance Rate Increases

Audit acceptance represents one of the most critical indicators of documentation quality. Across the pre-AI period, 62% of Local Files submitted by OMEGA were accepted without substantial clarification requests. While this rate is not atypical for large multinationals, the pattern of requests was becoming increasingly demanding, reflecting the global shift toward more assertive audit practices [20].

After implementing AI-enabled documentation, audit acceptance rose to 79%, an improvement of 27 percentage points. The increase is particularly notable in Romania, Poland, and Italy—jurisdictions where auditors routinely scrutinise functional analyses, benefit tests, and benchmarking coherence.

In Romania, for example, previous audits had identified narrative–numeric inconsistencies and insufficient justification for management service fees. The AI-generated documentation addressed these vulnerabilities automatically: functional analyses were verified against operational data, service descriptions were aligned with cost-centre outputs, and benchmarking results reflected up-to-date comparable financials.

This upward shift in audit acceptance places OMEGA above the median level reported in global transfer pricing surveys [21], suggesting that AI offers a measurable competitive advantage in audit preparedness.

4.3.4. Reduction in External Consultant Dependency

The quantitative evidence also highlights a structural shift in OMEGA’s resource allocation. Before the transition to AI, external consultants were responsible for portions of the documentation cycle—benchmarking refreshes, narrative drafting for certain jurisdictions, or reviewing inconsistencies flagged during internal control steps.

After AI deployment, dependency on consultancy services decreased by 41%, representing a substantial cost-efficiency gain. This figure does not indicate a diminished role for external expertise; rather, it reflects a more targeted use of such expertise for strategic advisory tasks instead of routine documentation. It also reduces the transfer of sensitive operational information outside the organisation, thereby enhancing governance integrity—an outcome consistent with the principles outlined in CSRD and ESG reporting frameworks [10,26].

4.3.5. Improvements in Benchmarking Currency and Alignment

Quantitative improvements are also evident in the frequency and reliability of benchmarking updates. Before automation, only about 60% of OMEGA’s benchmark analyses were refreshed annually, often due to time constraints or limited access to updated financial statements of comparables. In volatile industries, this lag created exposure to interquartile misalignment and increased audit risk.

Post-automation, benchmarking currency reached 100%, with automatic updates triggered when new data became available in Amadeus and Orbis [3,4]. The system also recalculated interquartile ranges for routine distribution activities, management services, and manufacturing support arrangements, ensuring alignment with OECD comparability standards [18].

This improved benchmark integrity translated into fewer deviations between actual profitability and arm’s-length ranges, a reduced probability of transfer pricing adjustments, and shorter internal review cycles.

4.3.6. Reduction in Inter-Document Variability Across Jurisdictions

Another set of quantitative improvements relates to the reduction in variability across Local Files. Prior to AI implementation, cross-jurisdictional divergence was significant; during review cycles, OMEGA identified up to 15–20% structural variation between documents prepared in different jurisdictions for subsidiaries performing nearly identical functions. This cross-country variability decreased to below 5% following AI integration.

While the numerical reduction reflects structural alignment, the underlying implication is more profound: documentation coherence has become institutionalised rather than dependent on individual drafting styles. This finding reinforces the argument advanced by Batrancea et al. [1], who highlight the importance of consistency in tax governance practices across global subsidiaries.

4.3.7. Time Savings in Internal Review and Approval Processes

The introduction of standardised structures, XAI layers, and real-time validation mechanisms also shortened internal approval cycles. Previously, multi-level reviews could extend across weeks—particularly when narrative sections required extensive rewriting or when benchmarking tables conflicted with ERP data.

After AI integration, the average internal review cycle decreased by 52%, supported by higher clarity, fewer inconsistencies, and increased transparency regarding the source and meaning of each data point. This reduction reflects the literature linking explainability to improved decision-making efficiency in corporate reporting contexts [22,23].

4.3.8. Environmental Impact Metrics and Sustainability Gains

Finally, the system generated quantifiable environmental benefits. OMEGA’s transition to digital documentation eliminated the need for paper-based reports, cross-border courier shipments, and physical archiving.

Environmental metrics confirm:

• a total reduction of 78 kg in annual paper consumption;
• approximately 2.3 tons of CO₂ emissions avoided due to the elimination of international document transport.

These results are directly aligned with the sustainability objectives articulated under the 2030 Agenda [24] and provide empirical support for the argument that digitalisation in the tax function contributes directly to environmental sustainability—an aspect increasingly relevant under CSRD requirements [10,26].

4.3.9. Consolidated Interpretation of Efficiency Gains

Taken together, the quantitative findings demonstrate that AI-enabled TP documentation supports a multidimensional improvement trajectory: higher operational efficiency, stronger audit defensibility, reduced compliance risk, increased sustainability performance, and enhanced internal governance. The magnitude and coherence of these results reflect not merely the impact of a digital tool, but the systemic integration of AI within OMEGA’s fiscal ecosystem—a conclusion consistent with emerging academic and professional research on the evolution of tax compliance functions [14,22,23].

4.3.10. Key Performance Indicators

To consolidate the quantitative results obtained across the OMEGA Group,

Table 1. Pre-AI vs. Post-AI performance indicators.

Indicator	2021 (Pre-AI)	2024 (Post-AI)	Change (%)
Avg. time to prepare local file	18 days	5.7 days	−68.3%
Documentation error rate	11.4%	2.1%	−81.5%
Avg. DPT adjustment (EUR)	€142,000	€81,500	−42.6%
Audit acceptance rate	74%	94%	+27%
AI compliance accuracy	–	96.2%	–

summarises the main performance indicators before and after the adoption of the AI-enabled documentation system. The data confirm the scale of operational, compliance, and governance improvements generated by the transition to automated and standardised TP workflows.

Additional operational gains include:

• A 58% reduction in audit clarification requests;
• A 23% reduction in external consulting fees;
• The introduction of automated monthly “compliance health reports” available across all jurisdictions.

These indicators reinforce the conclusion that AI produces improvements that extend beyond efficiency, contributing to more robust governance, strengthened audit readiness, and greater documentation consistency across the multinational structure.

4.4. Romania Transfers Pricing Adjustment Case Study

Romania provides a particularly instructive setting for examining the impact of AI-enabled documentation on audit exposure. The country’s transfer pricing framework is shaped by a combination of highly formalistic documentation requirements, an assertive audit environment, and a methodological stance that places significant weight on internal consistency and the demonstration of economic substance. Unlike some Western European jurisdictions, where audits tend to prioritise broad risk assessments, Romanian tax authorities frequently engage in detailed line-by-line reviews of Local Files, benchmarking studies, and intercompany service justifications. Within this context, the OMEGA case offers an opportunity to isolate the mechanisms through which AI tools contribute to audit resilience.

4.4.1. Structural Characteristics of the Romanian Audit Environment

Romania’s approach to transfer pricing audits has been extensively described as meticulous, documentation-driven, and strongly anchored in the presumption that inconsistencies indicate potential profit misalignment. Three structural features of the local audit environment are especially relevant. First, Romanian authorities apply a strict interpretation of functional analyses, expecting clear evidence linking operational roles to financial outcomes. Even small divergences between ERP-recorded activities and narrative descriptions tend to trigger inquiries. Second, the benefit test for intra-group services is interpreted expansively, requiring taxpayers not only to demonstrate the receipt of services but also to provide explicit documentation on how these services generated measurable improvements. Third, auditors place substantial emphasis on the annual renewal and methodological transparency of benchmarking studies, often requesting justification for the inclusion or exclusion of specific comparables. These expectations create a high compliance bar. In the years preceding AI adoption, OMEGA faced several rounds of clarifications, particularly concerning management services, cost allocation drivers, and deviations from interquartile benchmarks. Importantly, none of these issues originated from methodological flaws; instead, they were symptomatic of documentation misalignment—a problem that is common among multinationals operating in jurisdictions with demanding audit cultures [19,21].

4.4.2. Diagnostic Phase: Sources of Audit Exposure Prior to AI Integration

The baseline documentation submitted by OMEGA in Romania contained several vulnerabilities that became visible during the audit. The first set of issues related to the consistency between functional analysis and operational evidence. Although the subsidiary was characterised as a limited-risk distributor, certain managerial reports included references to local promotional tasks, which authorities interpreted as inconsistent with the expected risk profile. The discrepancy did not in itself invalidate the transaction structure; however, the absence of a clear explanation created an opening for auditors to scrutinise the margin positioning more aggressively.

A second vulnerability concerned the quality of the benefit test documentation. OMEGA provided detailed descriptions of service categories but lacked supporting operational evidence in specific areas. For instance, while strategic management services were adequately detailed, evidence relating to administrative support functions was fragmented across internal systems, complicating the audit trail. Given that the Romanian tax administration often challenges intra-group services on the grounds of duplication or perceived shareholder activities, incomplete documentation can significantly increase the likelihood of adjustments [18,19].

The third issue centred on the benchmarking set, which, although methodologically sound, relied on comparables whose financials had not yet been refreshed for the audited year. While this delay is common in practice, local auditors interpreted it as an indication that the taxpayer did not sufficiently validate the economic grounds of its transfer pricing policy. The absence of a contemporaneous refresh created a risk that interquartile ranges might not fully reflect market conditions [18].

4.4.3. AI-Supported Transfer Pricing Adjustment for the Romanian Subsidiary (2023)

To illustrate the operational impact of AI in a demanding audit jurisdiction,

Table 2. AI-supported DPT adjustment for Romanian subsidiary (2023).

Indicator	Value (EUR)
Net Revenue	12,400,000
Operating Expenses	11,620,000
Declared Operating Profit	780,000
Declared Operating Margin	6.29%
Benchmark Range (IQR)	7.5–12.8%
AI-Suggested Margin	9.3%
Adjusted Profit (DPT)	1,153,200
Resulting Adjustment	+373,200

summarises a representative transfer pricing adjustment identified for the Romanian subsidiary during the 2023 fiscal year. The case reflects how the AI system integrates benchmarking, ERP-derived financials, and explainability layers to substantiate arm’s-length outcomes.

The system identified that the declared margin fell below the arm’s-length interquartile range primarily due to the delayed recognition of intercompany management fees. By automatically recalculating profitability against updated comparables and aligning cost allocations with ERP evidence, the AI solution generated a contemporaneous TP file that was accepted by the Romanian tax authorities without additional clarification.

This example illustrates how AI strengthens audit resilience in jurisdictions with high formalism, where documentation coherence, benchmarking currency, and traceable benefit tests carry significant weight during inspection.

4.4.4. AI-Enabled Remediation: Structural Repairs Rather than Cosmetic Adjustments

AI integration produced a qualitative leap because the system did not simply correct individual inconsistencies—it restructured the logic through which evidence was assembled and connected. In the Romanian context, four AI-driven mechanisms proved decisive.

First, the NLP extraction layer systematically cross-referenced functional descriptions with operational data from ERP and management systems. When the AI model detected references to promotional or client engagement tasks, it automatically flagged discrepancies relative to the functional profile of a limited-risk distributor. These flags were not corrections but prompts for TP specialists to evaluate the appropriate classification. This process ensured that the final narrative contained a balanced and thoroughly reasoned explanation of the subsidiary’s operational scope, reducing ambiguity during audit review.

Second, the benefit test underwent a substantial transformation. AI cross-mapped every service category with internal evidence—emails, project logs, cost-centre records, and SOPs—bringing to the surface a level of granularity that had previously been difficult to consolidate manually. This granular documentation aligned closely with Romanian expectations, where auditors often ask taxpayers to provide “itemised evidence” of received services. The AI-generated structure did not merely present the services but contextualised them in relation to year-specific operational developments.

Third, the benchmarking upgrade mechanism ensured that the financial data of comparables was refreshed automatically as soon as updates became available in the Amadeus and Orbis databases [3,4]. This eliminated the temporal gap between the financial year-end and the update cycle that had previously created exposure. The AI system recalibrated the interquartile range for the tested party and linked profitability outcomes directly to ERP data, creating a clear and audit-ready alignment between method selection, data inputs, and final conclusions.

Fourth, the explainability layer (XAI) provided a transparent justification for every element of the Local File. In an environment where Romanian auditors frequently challenge documentation on the grounds of insufficient rationale, the ability to demonstrate the origin of each data point significantly enhanced the credibility of OMEGA’s submission. This aligns with recent policy discussions on responsible AI and its role in regulated environments [11,22,23].

4.4.5. Audit Outcome: Shifting the Narrative from Defence to Compliance Assurance

The audit outcome following AI implementation marked a clear departure from previous audit cycles. The documentation was accepted without adjustments, and clarification requests focused primarily on standard procedural questions rather than substantive challenges. This shift reflects not only the higher coherence of the Local File but also the enhanced capacity of the taxpayer to provide immediate explanations.

Importantly, the improved audit experience cannot be attributed solely to better narratives. The Romanian authorities paid particular attention to the congruence between profitability indicators and benchmark studies. Because the AI system refreshed data in real time and clearly documented any deviations or methodological changes, auditors encountered fewer ambiguities and were able to assess the economic substance more efficiently.

Moreover, the enhanced completeness of the benefit test made it possible for OMEGA to demonstrate the operational relevance of each service category. This aligns with the global trend emphasising substance over form in TP audits [18,19].

4.4.6. Governance Lessons from the Romanian Case

The Romanian case demonstrates that AI-enabled TP documentation is not simply a tool for producing more coherent reports; it is a structural instrument for embedding governance discipline. The system’s ability to detect inconsistencies, classify evidence, maintain contemporaneous documentation, and articulate reasoning in a transparent manner fosters a compliance culture that aligns with the expectations of demanding audit jurisdictions.

This case also illustrates the importance of integrating AI within a broader governance framework. Without managerial support, quality control processes, and a clear understanding of how evidence must be structured for audit scrutiny, the benefits of automation would be limited. The Romanian experience confirms that AI systems deliver the highest value when complemented by rigorous human review and a clear methodological strategy—an observation consistent with emerging literature on tax governance [1,18].

4.4.7. Consolidated Interpretation of Governance Effects

Overall, the Romanian case illustrates how AI addresses vulnerabilities that stem not from methodological flaws but from the complexity of assembling, aligning, and substantiating large volumes of evidence. By reconfiguring the data architecture, enriching benefit demonstration, maintaining benchmarking currency, and providing transparent justification layers, AI enables a more resilient audit posture in a jurisdiction characterised by stringent documentation standards.

This case provides empirical support for the broader argument that AI contributes not only to technical improvements but also to enhanced institutional robustness within TP documentation frameworks.

4.5. Organisational and Governance Effects

The organisational implications of introducing AI-enabled documentation at OMEGA extend well beyond efficiency gains. The transition has initiated a deeper reconfiguration of governance practices, decision-making structures, and the way the fiscal function interacts with other internal stakeholders. From a managerial standpoint, the most significant developments concern the consolidation of internal controls, the strengthening of accountability mechanisms, and the emergence of a more integrated compliance culture. These effects are neither incidental nor purely technological; they reflect the way AI reshapes institutional routines and encourages more disciplined documentation practices.

4.5.1. Shifts in Internal Coordination and Role Allocation

Before the implementation of AI, responsibilities for TP documentation were distributed across tax, finance, legal, and operational units, generating overlaps and inconsistencies. Local teams managed narrative drafting while the central TP group undertook benchmarking updates and final reviews. This division of labour, though common in multinationals, created structural blind spots: certain entities over-documented, others provided minimal detail, and reviewers spent significant time reconciling divergent interpretations.

The introduction of AI produced a more coherent allocation of responsibilities. With the core data extraction, structural drafting, and benchmarking validation handled by the system, the role of specialists shifted toward interpretative and evaluative functions. This phenomenon aligns with emerging research on the evolution of the accounting profession under automation, where expertise becomes increasingly focused on judgement-intensive activities rather than repetitive tasks [15]. Interviews conducted for this study confirm that TP experts at OMEGA now devote a greater proportion of their time to analysing year-on-year profitability trends, assessing the strategic relevance of intercompany services, and evaluating changes in the economic environment.

4.5.2. Enhancement of Accountability and Transparency

A defining feature of the new governance structure is the heightened transparency generated by the system’s traceability functions. The explainability layer (XAI) documents the source, meaning, and evidentiary basis of each paragraph, table, or calculation included in the Local File. In practice, this means that managers can audit the internal logic of the documentation without relying on fragmented email exchanges or verbal explanations.

This traceability has strengthened accountability mechanisms at multiple levels. Local teams are now responsible for validating the operational accuracy of AI-generated narratives, while the central TP function oversees methodological coherence. The separation between procedural compliance and economic interpretation has become clearer. This reflects a broader global trend in tax governance, where documentation integrity and process auditability are increasingly regarded as indicators of responsible corporate conduct [18,21].

4.5.3. Cultural Adaptation and Professional Behavioural Change

AI implementation also generated notable behavioural effects. Initial resistance—predominantly based on concerns regarding loss of editorial autonomy—gradually diminished as employees interacted with the system and observed its level of precision. The system’s ability to flag inconsistencies early in the drafting process enabled teams to build trust in its reliability. Over time, routines stabilised, and the system became a reference point for internal discussions, reducing subjective differences in interpretation.

The cultural adaptation process described by OMEGA staff is consistent with wider observations in the literature: effective AI integration often produces a hybrid work environment in which employees rely on automated structure while retaining full authority over final judgement [5]. This hybrid model proved particularly valuable in jurisdictions with demanding audits—Romania being a leading example—where TP experts must articulate complex rationales under tight procedural scrutiny.

4.5.4. Strengthening the Organisation’s Capacity for Audit Preparedness

One of the most visible governance outcomes concerns audit preparedness. Because the AI system continuously monitors data completeness, alignment between Master File and Local Files, and the currency of benchmarking studies, OMEGA has moved from a reactive audit posture to a proactive compliance assurance model. The compliance dashboard flags risks before auditors identify them, enabling internal teams to address inconsistencies well in advance of potential reviews.

This proactive stance is strongly aligned with the compliance transformation advocated in OECD’s digital tax administration recommendations [20]. Instead of focusing on documentation production at year-end, OMEGA now maintains a contemporaneous documentation environment, which significantly reduces both audit exposure and the internal stress traditionally associated with TP reviews.

4.5.5. Consolidated Interpretation of Sustainability Outcomes

Overall, the organisational and governance effects observed at OMEGA reveal that AI serves as a structural catalyst, reinforcing internal controls, facilitating interdepartmental coordination, and elevating the maturity level of the fiscal function. Rather than replacing human judgement, AI has created a framework in which judgement can be exercised more effectively, consistently, and transparently.

These changes reflect not only compliance optimisation but also a deeper institutional evolution toward disciplined, accountable, and sustainability-aligned tax governance [1,26].

4.6. Sustainability Outcomes

The sustainability implications of AI-enabled transfer pricing documentation extend beyond efficiency gains and operational streamlining. At OMEGA, the integration of AI into the fiscal function produced measurable environmental outcomes, improved governance transparency, and contributed to a broader organisational shift toward responsible reporting practices. These developments are consistent with the direction of global regulatory frameworks—most notably the Corporate Sustainability Reporting Directive (CSRD) and Agenda 2030—and reflect the growing expectation that tax governance serves not only compliance objectives but also the long-term sustainability commitments of multinational enterprises.

4.6.1. Environmental Footprint Reduction

One of the most immediate and quantifiable sustainability effects relates to the reduction in material resource consumption. Prior to AI implementation, OMEGA routinely printed and circulated transfer pricing documentation across jurisdictions for internal review and audit preparation. Despite internal digitalisation efforts, physical copies were still required in several cases due to procedural habits or auditor requests. AI integration allowed the group to eliminate these practices entirely. System-generated Local Files, benchmarking tables, and audit packages are now consolidated and exchanged exclusively through secure digital channels.

Empirical data indicate a 78 kg annual reduction in paper consumption, alongside the elimination of approximately 2.3 tons of CO₂ emissions associated with cross-border shipment of documentation. Although these figures may appear modest relative to the scale of OMEGA’s global operations, they represent a structural transition toward low-impact compliance processes and align with findings from sustainability literature showing that administrative digitalisation can materially reduce environmental footprints in knowledge-intensive sectors [16,25].

4.6.2. Contributions to Governance Transparency and Responsible Reporting

Beyond environmental metrics, the AI system contributes meaningfully to the governance dimension of sustainability. CSRD introduces more stringent expectations regarding transparency, internal control, and the traceability of non-financial information. Transfer pricing, due to its reliance on high volumes of qualitative and quantitative evidence, represents an area where governance weaknesses can undermine credibility. The explainability and traceability features of OMEGA’s AI solution reinforce internal control mechanisms by providing audit-ready documentation of sources, assumptions, and analytical steps.

This increased transparency supports responsible tax behaviour, a theme highlighted in recent international guidelines and legislative debates [8,14]. In practice, this means that senior management and sustainability teams can rely on structured evidence trails when preparing CSRD-aligned disclosures concerning tax governance, risk management, and compliance integrity. Given the rising scrutiny of corporate tax behaviour within ESG assessments, the AI-enabled documentation system strengthens OMEGA’s capacity to demonstrate responsible stewardship of its fiscal obligations.

4.6.3. Integration of the Fiscal Function into the Sustainability Strategy

A broader consequence of the AI transformation is the repositioning of the fiscal function within OMEGA’s organisational sustainability architecture. Historically, tax departments operated in parallel to ESG reporting units, with limited integration. However, the introduction of transparent, data-driven TP documentation has facilitated closer collaboration between tax, finance, and sustainability teams. In interviews, several managers noted that the availability of traceable evidence and the reduction in operational noise enabled more substantive discussions on how fiscal processes contribute to the organisation’s long-term sustainability objectives.

This reflects an emerging trend in corporate governance scholarship suggesting that tax compliance, when supported by robust digital infrastructures, can evolve into a vector for broader sustainability outcomes [16,25]. AI thus serves as a bridge between compliance and sustainability, reinforcing the organisation’s ability to demonstrate responsible conduct and to integrate fiscal considerations into ESG planning cycles.

4.6.4. Consolidated Interpretation of Audit and Compliance Implications

Overall, the sustainability gains at OMEGA are not limited to environmental or procedural improvements; they reflect a deeper institutionalisation of responsible governance within the transfer pricing function.

By reducing the environmental footprint, enhancing documentation transparency, and strengthening cross-functional collaboration, AI-enabled documentation provides a model for how digital tools can advance sustainability in highly regulated corporate domains.

These effects position the fiscal function as a contributor—not merely a respondent—to the organisation’s broader sustainability commitments [10,16,25].

5. Discussion

The findings of this research must be interpreted within the broader landscape of digital transformation in tax administration and the evolving expectations placed on multinational enterprises with respect to transparency, internal control, and sustainability. While Section 4 presented empirical evidence from the OMEGA case, the present discussion situates these results within current academic debates and regulatory developments, highlighting their conceptual significance and practical implications. Three overarching themes emerge: the redefinition of compliance through digital infrastructures, the interplay between AI-driven documentation and tax governance, and the integration of sustainability considerations into fiscal processes.

5.1. Digital Infrastructures and the Redefinition of Compliance

One of the central insights of this study is that AI-enabled documentation systems do not merely accelerate existing compliance processes; they transform the underlying architecture of how compliance is produced. The shift mirrors a wider movement observed across tax administrations, where digitalisation increasingly becomes a condition for both regulatory enforcement and corporate readiness [18,19]. In this context, the reduction in documentation time at OMEGA should not be understood simply as an operational gain, but as a structural reconfiguration of how information flows through the organisation.

The literature on automation in accounting suggests that structural benefits arise when AI displaces repetitive tasks and enables specialists to concentrate on interpretation, negotiation, and strategic risk assessment [15]. The empirical findings from OMEGA confirm this pattern: employees reported that the elimination of manual drafting allowed them to focus on substantive issues such as profitability trends, intercompany rationale, and methodological justification. This evolution implies a redefinition of the skill profile required within the fiscal function, where expertise becomes increasingly centred on evaluative and analytic capacities rather than clerical proficiency.

Moreover, the coherence introduced by the AI system suggests that digital infrastructures may help overcome persistent coordination failures that often arise in complex multinational groups. The standardisation of narratives, the systematic validation of benchmarking data, and the alignment between local and group-level documentation are not merely process improvements; they address structural challenges that traditional governance mechanisms struggled to resolve. In this respect, OMEGA’s experience resonates with the broader argument that digitalisation, when properly embedded, enhances the organisational capacity to fulfil regulatory obligations in a consistent and predictable manner [18,19].

5.2. AI, Governance Discipline, and the Evolution of Tax Control Frameworks

A second theme concerns the interaction between AI-driven documentation and governance integrity. OMEGA’s experience supports the claim that explainability, traceability, and contemporaneous documentation—elements embedded within AI workflows—serve as proxies for governance discipline. This finding aligns with the direction of major international regulatory frameworks. Both CSRD [10] and emerging AI governance standards [11,14] place renewed emphasis on auditability, internal control, and the ability to justify procedural decisions.

From this perspective, the enhanced audit acceptance observed in the Romanian case reflects not only improved documentation quality but also a shift in the organisation’s governance posture. Auditors increasingly evaluate not merely the correctness of financial data but the robustness of internal processes through which that data is produced. The AI system provided OMEGA with an institutional structure capable of demonstrating procedural discipline—an aspect that is becoming central to modern tax control frameworks [18,19].

This aligns with a broader conceptual shift in the literature, which argues that tax compliance is no longer solely a matter of technical calculation but also a reflection of governance quality [1]. Under this view, the fiscal function is an integral component of the organisation’s internal control system, and demonstrating responsible tax behaviour requires transparent documentation protocols, consistent methodological application, and clear evidence trails. The OMEGA case illustrates how AI can operationalise these expectations by embedding governance logic into routine compliance activities.

5.3. AI as a Catalyst for Cross-Functional Integration

The empirical data also reveal a noteworthy organisational evolution: the fiscal function becomes more tightly integrated with other corporate units. The increased visibility offered by compliance dashboards and XAI layers enables cross-functional collaboration in areas that historically operated in parallel—finance, operations, legal, and sustainability reporting.

This integration has two important implications. First, it enhances the organisation’s capacity to maintain alignment between financial outcomes, operational realities, and contractual structures—an alignment that is central to the integrity of TP documentation [1]. Second, it encourages a more holistic view of corporate governance in which fiscal processes are no longer isolated but contribute actively to risk management, strategic planning, and sustainability objectives.

The alignment with ESG agendas is particularly relevant. As reporting expectations intensify, tax governance becomes an integral part of sustainability narratives, especially in the governance pillar [10,16]. By enabling transparency and reducing inconsistencies, AI-generated documentation helps organisations demonstrate responsible conduct in areas historically susceptible to opacity. This observation supports emerging academic perspectives suggesting that digital infrastructures can serve as agents of institutional coherence across sustainability, finance, and compliance domains [7,18].

5.4. Sustainability Implications Beyond Environmental Gains

While environmental reductions in paper and CO₂ emissions are important, the OMEGA case indicates that the sustainability relevance of AI-enabled documentation lies primarily in its governance contributions. Responsible tax behaviour, transparency in transfer pricing, and consistent application of OECD principles reinforce the legitimacy of corporate tax practices—an issue increasingly scrutinised by investors, regulators, and civil society [16,25].

The fact that OMEGA can now provide contemporaneous, traceable evidence for its TP positions supports the call for increased accountability in global taxation frameworks. As multinational enterprises face rising expectations for disclosure under CSRD, having a robust digital infrastructure is not merely beneficial but necessary. This insight suggests that sustainability-oriented tax governance will increasingly depend on the integration of AI-driven documentation tools, especially in complex groups.

5.5. Limitations and Directions for Future Research

Although the OMEGA case provides a rich empirical foundation, certain limitations must be acknowledged. First, the findings derive from a single multinational operating in a relatively advanced digital environment. Organisations with lower digital maturity may require different implementation pathways or experience slower adoption curves. Second, the analysis does not explore the implications of AI for dispute resolution mechanisms such as MAP or APA procedures—areas where future research could provide valuable insights. Third, while the system clearly enhances audit defensibility, its interaction with tax administrations adopting their own AI tools remains an open question, inviting further investigation.

5.6. Overall Consolidated Interpretation of Findings

Taken together, the discussion demonstrates that AI-enabled transfer pricing documentation represents not a technological add-on but a systemic innovation with implications across compliance, governance, and sustainability. The OMA case supports the proposition that digital infrastructures strengthen the coherence, auditability, and strategic value of the fiscal function.

As regulatory frameworks continue to evolve, organisations that integrate AI into their tax control frameworks will likely be better positioned to meet the growing expectations for transparency, responsibility, and sustainability.

6. Conclusions

This study examined the integration of AI-enabled documentation within the transfer pricing function of a multinational enterprise and showed how digital infrastructures can reshape compliance processes, governance practices, and sustainability outcomes. Evidence from the OMEGA case indicates that AI delivers more than operational efficiencies: it changes how TP documentation is produced, validated, and assessed by internal and external stakeholders, consistent with wider digitalisation trends reported in recent research [19,20,21].

A central conclusion is that AI strengthens the structural integrity of transfer pricing documentation by embedding consistency, contemporaneity, and methodological transparency into the workflow. These features address long-standing vulnerabilities of manual practices, including fragmented data sources, narrative–numeric mismatches, outdated benchmarking sets, and inconsistent cross-jurisdictional reporting. By systematising these processes, AI helps organisations align more closely with expectations in the OECD Transfer Pricing Guidelines [18] and emerging international frameworks on responsible tax governance [1].

The Romanian case study highlights the substantive impact of AI on audit preparedness. In jurisdictions with a strict, documentation-driven audit culture, the system’s ability to trace evidentiary links, refresh comparables in a timely manner, and articulate the operational substance behind intercompany services can materially reduce exposure to adjustments. This reinforces the view that AI’s value lies not only in automation but also in embedding governance logic into daily compliance practices [12,19].

Digitalisation also has broader organisational implications. AI shifts the fiscal function away from clerical activities and towards interpretative, analytical, and strategic tasks. This transition can strengthen cross-functional collaboration and increase the fiscal function’s contribution to corporate governance, consistent with prior work on how automation changes professional roles in accounting and tax [7,15]. Integrating explainability and traceability features further supports managerial oversight by improving visibility into the documentation process, consistent with principles reflected in ISO/IEC 42001 [10] and the EU AI Act [11].

Finally, the study provides evidence that AI-enabled TP documentation can support corporate sustainability, both through measurable resource savings and through improved governance transparency. These developments align with the direction of European sustainability frameworks, particularly CSRD, which emphasise internal control, traceability, and responsible tax behaviour as part of the governance pillar of ESG reporting [10,16,26].

7. Patents

No patents have been generated or submitted as a result of the work reported in this manuscript.