ESG Meets AI : When Algorithms Write the Sustainability Report

The Pressure Is Mounting

The Corporate Sustainability Reporting Directive (CSRD) presents thousands of European companies with an enormous challenge: for the first time, they must report comprehensively and in a standardised manner on environmental, social and governance aspects. Reporting under the European Sustainability Reporting Standards (ESRS) requires the collection of hundreds of data points from the most diverse areas of the business -- from CO₂ emissions and supply chain information to biodiversity impacts.

Given this enormous effort, more and more companies are turning to artificial intelligence as a tool for sustainability reporting. AI promises efficiency, scalability and consistency. Yet the use of algorithms in this highly regulated field raises fundamental questions: where does meaningful automation end -- and where does greenwashing begin?

The CSRD at a Glance

Scope

The CSRD (Directive (EU) 2022/2464) significantly expands the range of companies subject to reporting obligations:

• Large public-interest entities: Already subject to reporting obligations from the 2024 financial year
• All large companies (two of three criteria: > 250 employees, > EUR 50 million turnover, > EUR 25 million total assets): From the 2025 financial year
• Listed SMEs: From the 2026 financial year (with a transitional period until 2028)
• Non-EU companies with EU net turnover exceeding EUR 150 million: From the 2028 financial year

Reporting Standards: The ESRS

The European Sustainability Reporting Standards (ESRS) were developed by the European Financial Reporting Advisory Group (EFRAG) and adopted by the European Commission as delegated acts on 31 July 2023. They comprise:

• ESRS 1 (General Requirements) and ESRS 2 (General Disclosures): Mandatory for all companies
• ESRS E1-E5 (Environment): Climate change, pollution, water and marine resources, biodiversity, circular economy
• ESRS S1-S4 (Social): Own workforce, workers in the value chain, affected communities, consumers and end users
• ESRS G1 (Governance): Business conduct

The Principle of Double Materiality

A core principle of the CSRD is double materiality. Companies must report both on how sustainability matters affect the company financially (financial materiality) and on how the company itself impacts the environment and society (impact materiality).

Where AI Can Help in ESG Reporting

1. Automated Data Collection and Aggregation

The greatest challenge of CSRD reporting lies in data procurement. Companies must consolidate information from a multitude of internal and external sources: ERP systems, HR platforms, energy management systems, supplier portals and many more.

AI-powered systems can significantly accelerate this data collection:

• Automatic extraction of ESG-relevant data from unstructured documents (invoices, supplier reports, certificates)
• Real-time monitoring of energy consumption and emissions via IoT sensors
• Consolidation of data from different locations and subsidiaries

2. Materiality Assessment

The double materiality assessment requires the systematic evaluation of a large number of sustainability topics. AI can provide support through:

• Natural Language Processing (NLP) for analysing stakeholder feedback, media coverage and regulatory developments
• Sentiment analysis for identifying emerging ESG risks
• Cluster analyses for grouping and prioritising materiality topics

3. Carbon Footprint Calculation

The calculation of Scope 3 emissions -- that is, indirect emissions in the upstream and downstream value chain -- presents particular challenges. AI models can:

• Automatically assign emission factors from databases
• Fill gaps in primary data through model-based estimates
• Calculate scenario analyses for different reduction pathways

4. Report Drafting and Text Generation

Large Language Models (LLMs) can support text creation:

• Drafting of standard passages and boilerplate text
• Translation of reports into multiple languages
• Consistency checks across different report sections

The Limits of AI: Where It Gets Dangerous

Hallucinations and Data Quality

AI systems, particularly Large Language Models, are susceptible to hallucinations -- the generation of plausible-sounding but factually incorrect statements. In a sustainability report that is subject to audit requirements, this can have fatal consequences:

• Fabricated emission values or reduction targets
• False references to norms or standards
• Inconsistent information between different report sections

Greenwashing by Algorithm

The uncritical use of AI in ESG reporting can lead to a new form of greenwashing. When algorithms are trained to package company data into the most positive narrative possible, there is a risk that:

• Negative developments are obscured through skilful phrasing
• Data gaps are silently filled with optimistic assumptions
• Complex relationships are impermissibly oversimplified

The European legislator is taking greenwashing increasingly seriously. The European Commission presented a proposal for a Green Claims Directive in March 2023, which would have subjected environmental claims to verification requirements. Although the current status of the proposal is uncertain, the direction is clear: unsubstantiated or misleading sustainability claims will be sanctioned more strictly in the future.

The Black Box Problem

Many AI models operate as a black box: their decision-making processes are opaque to outsiders -- and often to the companies using them as well. This conflicts with the transparency requirements of the CSRD and the ESRS, which demand a comprehensible presentation of methods and assumptions.

Audit Obligations and Verification

Obligation for Limited Assurance

The CSRD provides that sustainability reports must be subjected to limited assurance engagement by an independent auditor or assurance provider. In the medium term, an upgrade to reasonable assurance is planned.

What Does This Mean for AI-Supported Reports?

Auditors must be able to trace and verify the information presented in the report. With AI-supported reporting, particular challenges arise:

• Auditability: The auditor must understand the methodology of the AI systems and be able to assess the plausibility of the results
• Data provenance: Every data point used in the report must be traceable to its primary source
• Model validation: The AI models used must be regularly checked for accuracy and reliability
• Bias control: Systematic distortions in the training data can lead to systematically incorrect reporting results

The Role of the DRSC

The German Accounting Standards Committee (DRSC) plays a central role in implementing the ESRS in Germany. It represents German interests in the development of European standards and supports practical implementation. Companies should closely follow the DRSC's pronouncements, as they provide valuable guidance on the interpretation of the standards.

The EU Taxonomy as a Framework

Alongside the CSRD, the EU Taxonomy Regulation (EU) 2020/852 forms another important building block of the European sustainability framework. It defines which economic activities are to be classified as environmentally sustainable and requires affected companies to disclose taxonomy metrics (turnover, capital expenditure, operating expenditure).

AI can also assist with taxonomy classification -- for instance, in the automated mapping of economic activities to the taxonomy's six environmental objectives. However, assessing whether an activity makes a substantial contribution and causes no significant harm (Do No Significant Harm -- DNSH) frequently requires qualitative judgements that are beyond a purely algorithmic solution.

Practical AI Tools for ESG Reporting

The market for AI-powered ESG software is growing rapidly. Relevant solution categories include:

Data Management Platforms

• Automated collection and consolidation of ESG data from various sources
• Integration with existing ERP and HR systems
• Real-time dashboards for sustainability performance

Emissions Management

• Automatic calculation of Scope 1, Scope 2 and Scope 3 emissions
• Modelling of reduction scenarios
• Benchmarking against industry values

Report Generation Tools

• Automatic generation of draft reports following the ESRS structure
• Cross-referencing between different standards (ESRS, GRI, TCFD)
• Gap analysis to identify missing data points

Risk Analysis

• AI-powered monitoring of ESG risks in the supply chain
• Early warning systems for regulatory changes
• Automated media analysis for reputation risks

The Human Oversight Imperative

Despite all the potential of AI, one immutable principle applies: Responsibility for the sustainability report lies with the company and its governing bodies -- not with the algorithm.

Minimum Requirements for Human Oversight

Companies that use AI in ESG reporting must ensure:

• Final decision by humans: All AI-generated content must be reviewed and approved by qualified professionals
• Methodological transparency: The AI systems used and their functioning must be disclosed in the report
• Quality assurance: Systematic controls of AI results through sampling and plausibility checks
• Competence building: Employees must be able to critically question and correct AI results
• Documentation: The entire process -- from data input through AI processing to human validation -- must be documented without gaps

Governance Framework for AI in ESG Reporting

A robust governance framework should include:

• AI policy: Clear rules for the use of AI in reporting
• Role allocation: Definition of which tasks AI may perform and which require human expertise
• Escalation mechanisms: Procedures for dealing with divergent or implausible AI results
• Regular audits: Review of AI systems by internal or external experts

Practical Recommendations

Short-term

1. Take stock: What ESG data already exists? Where are the gaps?
2. Start pilot projects: Initially deploy AI for clearly defined tasks (e.g. data collection, emissions calculation)
3. Involve the auditor early: Inform the auditor about AI usage already in the planning phase

Medium-term

1. Build data infrastructure: Implement a central ESG data platform
2. Standardise processes: Define clear workflows for AI-supported reporting
3. Build competence: Training for sustainability, IT and finance teams

Long-term

1. Integration: Integrate ESG reporting into regular financial reporting
2. Continuous monitoring: Move from annual reporting to continuous sustainability management
3. Best practices: Share experiences and help shape industry standards

Conclusion

AI is a powerful tool for ESG reporting -- but no substitute for human responsibility and professional expertise. The regulatory framework of the CSRD, ESRS and EU Taxonomy demands precision, transparency and traceability. Those who deploy AI wisely can manage the enormous effort of sustainability reporting. Those who adopt it uncritically risk erroneous reports, audit failures and accusations of greenwashing.

At compleneo, we support you with CSRD-compliant sustainability reporting, the implementation of suitable AI tools and ensuring the auditability of your reports. Get in touch with us.