Key findings highlight that most corporate carbon-emissions reporting is inconsistent and lacks critical information, with around 40% of reported Scope 1 and 2 emissions not specifying their scope or operational coverage. Institutional investors require reliable, comparable, and transparent climate data for investment decisions and portfolio-emissions reporting. While AI has advanced in extracting corporate climate disclosures, transforming this data into comparable format remains challenging due to substantial human judgment and effective human-AI collaboration needed.
Challenges in reported emissions comparability
- Operational coverage: Approximately 36% of Scope 1 and Scope 2 emissions disclosures do not specify whether they cover a company's entire operations or only select business activities, leading to potential analysis errors if assumed full coverage.
- Decarbonization-target coverage: Around 56% of corporate climate targets analyzed did not explicitly state the share of emissions they cover, with critical frameworks like the Science-Based Targets initiative (SBTi) requiring near-term Scope 3 targets to cover at least 67% of emissions and most long-term targets at least 90%.
- Scope 2 reporting method: Nearly half (47%) of reported Scope 2 emissions do not specify whether they use a location-based or market-based approach, which can significantly impact reported figures.
- Scope 1 and Scope 2 totals divergence: 8.7% of disclosures showed a reported combined Scope 1 and Scope 2 figure not equal to the sum of individually reported values, reflecting differing carbon-accounting approaches.
Human judgment and verification
- Over 200 types of judgment calls are needed to address inconsistent, incomplete, or ambiguous corporate disclosures, as reported climate disclosures must align with broader company data and activities (e.g., production volumes).
- For example, a carmaker reporting exceptionally low Scope 2 emissions may exclude emissions from subsidiaries using an equity-share approach, which is valid under the GHG Protocol but produces materially different totals compared to an operational-control approach.
AI and structured data for investment-ready outputs
- Metadata, quality checks, and interpretation are created independently of data extraction to ensure models and investment analytics produce consistent, comparable results.
- For instance, projecting future emissions requires resolving metadata questions such as whether reported emissions and targets use the same accounting basis and whether values are complete and comparable.
- MSCI's models leverage this structured data foundation to deliver consistent projection estimates across companies.
Conclusion
While AI accelerates climate disclosure collection and processing, corporate reporting inconsistencies necessitate expert judgment to transform data into investment-grade quality. Reliable portfolio-level emissions reporting, driven by expanding disclosure and assurance requirements (e.g., in the EU), underscores the need for structured methodologies and governance. The optimal approach combines AI efficiency with human expertise to ensure data accuracy and comparability for real-world investment decisions.
