Global climate reporting at risk as AI falls short in value-chain emissions measurement

According to the review, AI is most commonly used for four functions: integrating data from enterprise systems, scraping emission-factor databases, mapping supplier information to reporting categories and identifying anomalies in reported data. In these settings, AI reduces manual workload, speeds up processing and helps standardize inconsistent inputs.

A new academic review warns that the global push for corporate climate transparency is being undermined by structural weaknesses in how organizations measure and assure Scope 3 greenhouse gas emissions. Their analysis raises concerns that despite rapid adoption of artificial intelligence to automate sustainability reporting, AI tools remain unable to overcome the deep systemic, political and data-quality problems that define value-chain emissions accounting.

The paper, “Conceptualization of Artificial Intelligence Use for GHG Scope 3 Emissions Measurement, Reporting, Monitoring, and Assurance: A Critical Systems Perspective,” published in Sustainability, evaluates corporate reporting obligations under the IFRS S2 Climate-related Disclosures and the Greenhouse Gas Protocol’s Scope 3 Standard. The authors apply a critical systems thinking lens to examine where AI can meaningfully support carbon accounting and where it creates new risks or reinforces hidden weaknesses across supply chains.

Their findings present a clear message: companies are turning to AI platforms to streamline climate reporting, but even the most advanced tools cannot resolve the underlying fragmentation, inadequate data infrastructure and complex power dynamics that make Scope 3 emissions the most difficult, and often the least credible, component of corporate climate disclosures.

How Scope 3 emissions became the most complex challenge for corporate climate reporting

The study examines why Scope 3 emissions account for the majority of corporate climate impact yet remain the least reliable in disclosures. The authors point out that Scope 3 covers 15 categories of upstream and downstream activities, from purchased goods and capital equipment to product use, waste and end-of-life management. In most sectors, these emissions exceed those of direct operations by a large margin.

Despite this importance, companies struggle to comply with the Scope 3 Standard because accurate value-chain information depends on thousands of suppliers, contractors, distributors and customers. Many suppliers lack the resources, expertise or incentives to collect primary emissions data. As a result, companies rely heavily on secondary sources, industry averages and proxy datasets that dilute accuracy and create wide reporting variability across organizations.

The authors describe this reliance on secondary data as a systemic problem. Without robust value-chain engagement and verifiable primary data, companies tend to compile emissions inventories that reflect assumptions, approximations and methodological inconsistencies rather than real carbon flows. In developing and emerging economies, where a large proportion of global manufacturing occurs, data gaps are particularly severe.

From an assurance perspective, the authors argue that these weaknesses create substantial risk. Assurance engagements for Scope 3 data remain largely voluntary, underdeveloped and dependent on checklists or procedural reviews that lack depth. Limited professional expertise in carbon accounting and unclear boundary-setting practices compound the problem, making Scope 3 assurance the least mature component of climate verification.

The paper also sheds light on the critical importance of boundary judgments and materiality decisions that shape what companies include or exclude in their claims. These decisions carry political and ethical dimensions that go beyond technical calculations, yet they are often made without explicit scrutiny, leaving reporting vulnerable to bias and greenwashing.

Where AI helps and where it cannot replace human judgment

Next up, the study explores how AI is currently integrated into sustainability reporting and the realistic limits of what it can deliver. The authors analyze capabilities marketed by leading ESG and sustainability platforms including Sustain.Life, Cority, Diligent, Osapiens, SAP Sustainability Tower, Sphera, Novisto, Workiva, Key ESG and others.

According to the review, AI is most commonly used for four functions: integrating data from enterprise systems, scraping emission-factor databases, mapping supplier information to reporting categories and identifying anomalies in reported data. In these settings, AI reduces manual workload, speeds up processing and helps standardize inconsistent inputs.

However, the authors emphasize that these functions are largely administrative and do not solve the core problems that limit the accuracy of Scope 3 emissions. AI tools depend almost entirely on secondary emissions factors, proxy datasets and supplier information that may be incomplete, outdated or inaccurate. When primary data are missing, AI can only make inferences based on imperfect proxies. As a result, automation does not meaningfully improve the underlying precision of reporting.

The authors further argue that AI does not offer the critical reflection required for Scope 3 assurance. Anomaly detection algorithms flag outliers but cannot judge materiality, interrogate management assumptions, examine power relationships within supply chains or challenge boundary choices that shape reported emissions. These tasks require human judgment, professional skepticism and a systems-based understanding of how corporate structures influence climate disclosures.

The paper also identifies risks associated with increasing reliance on AI. Because many AI models operate as black boxes, assurance professionals cannot always evaluate how conclusions are generated. This complicates verification and introduces a new layer of opacity into reporting. Additionally, the authors note that AI-driven reporting systems often prioritize efficiency and scale, which can lead to mechanistic outputs that overlook stakeholder context or reduce sensitive value-chain issues to numerical abstractions.

Another critical issue raised by the authors is the environmental footprint of AI itself. Large language models, cloud computing infrastructures and data-center operations contribute to upstream Scope 3 emissions for companies that use them. The study warns that organizations adopting AI-powered reporting solutions must also recognize the emissions associated with the technology. Without doing so, they risk shifting carbon burdens rather than eliminating them.

Why a critical systems approach is essential for reliable Scope 3 assurance

The study also explores the need for a critical systems perspective to improve the credibility of Scope 3 reporting. The authors argue that traditional “hard systems” approaches, focused on quantification, modeling and optimization, cannot address the broader social, political and ethical issues embedded in value-chain emissions accounting.

Critical systems thinking (CST), as applied by the authors, centers on three principles: critical awareness, emancipation and methodological pluralism. These principles call for a deeper examination of the assumptions, power structures and stakeholder relationships that influence how Scope 3 emissions are calculated, reported and audited.

According to the review, two power asymmetries dominate Scope 3 reporting. The first is the imbalance between large corporations and smaller suppliers. Many suppliers lack the resources to measure emissions, while large reporting entities rely on them for data quality. This structural imbalance can lead to incomplete inventories and pressure on suppliers to comply without adequate support.

The second asymmetry exists between corporate management and assurance providers. While auditors must challenge management claims to ensure credible reporting, limited expertise in Scope 3 methodologies, combined with the complexity of supply chains, reduces the effectiveness of assurance procedures. AI does not eliminate these tensions; in some cases, it may reinforce them by providing a veneer of precision that masks deep data gaps.

The authors call for greater government involvement in coordinating standardized Scope 3 data collection and ensuring that national and international bodies share responsibility for emissions verification. Without stronger public-sector leadership, they argue, corporate reporting will remain fragmented and overly dependent on voluntary initiatives.

The authors also point to the need for a shift in professional training for auditors and sustainability practitioners. Critical systems competencies, such as boundary critique, stakeholder analysis and reflective judgment, must be integrated into assurance frameworks. AI should be treated as a tool that supplements, rather than replaces, these human capabilities.

While AI has clear potential to support emissions reporting, it must be used in conjunction with systems-based governance structures that emphasize ethical analysis, data quality, human oversight and equitable value-chain engagement. Without these elements, Scope 3 reporting will continue to suffer from uncertainty and inconsistency, regardless of technological advancements.