How C-Suite Executives Can Use AI to Identify and Prioritize Material ESG Issues

Artificial intelligence gives executive teams a systematic way to identify, analyze, and prioritize the ESG issues that genuinely matter to their business—provided the process is structured deliberately. The practical approach that emerges from academic research, vendor implementations, and a corporate case study is a four-stage framework: aggregate and weight diverse data sources, anchor the analysis in an industry benchmark, apply semantic analysis rather than keyword counting, and let humans make the final stakeholder-specific prioritization. This article lays out that framework for a leadership audience; it is a strategic roadmap, not a technical guide, and it does not recommend specific vendors or interpret disclosure law.

The Strategic Imperative: Why AI for ESG Materiality?

Traditional materiality assessments struggle with volume and ambiguity. Companies emit enormous quantities of unstructured text—filings, sustainability reports, news coverage, social media commentary—and what stakeholders consider material can look very different depending on the stakeholder group and can shift quickly, as the team behind C3 AI's ESG program observed while developing machine-learning models that scan publicly available text near-continuously to detect which issues are being discussed and how material they are.

The core problem with manual or simplistic methods is that they measure the wrong things. Siyu Shen, a graduate student at Kyushu University's Graduate School of Economics and first author of a study on AI-assisted ESG disclosure analysis, puts it directly: "High-quality ESG disclosure helps stakeholders assess business performance and risks. Yet counting mentions, or judging whether the language sounds positive or negative, tells us little about how seriously a company treats those issues." For a board, that distinction is the difference between a report that flatters and one that informs.

AI addresses the gap on two fronts. It automates collection and initial topic identification across thousands of documents, and it brings analytical depth—through natural language processing—to the substance of what is disclosed. The result is an assessment that is faster, more comprehensive, and more reproducible than manual review, while still leaving the consequential judgment calls to executives.

Stage 1: Comprehensive Data Aggregation and Source Weighting

The first stage is gathering inputs, and the discipline here is weighting. Not all sources deserve equal influence. In Baker Hughes' AI-assisted materiality assessment, described by team member Caekebeke in an account reported by Constellation Research, filings with the Securities and Exchange Commission, sustainability reports, and annual reports carried a higher weighting than social media, where "everyone is a sustainability expert." The project also incorporated reports from customers, competitors, investors, and NGOs.

The weighting decision matters because it encodes credibility into the analysis before any model runs. A regulator-facing filing and an anonymous social post may discuss the same issue, but they represent very different evidentiary value. Executives should require that their teams define explicit weighting criteria—source authority, verifiability, and relevance to the company's sector—rather than letting the tool treat all text as equal.

Speed is a genuine benefit at this stage. In the Baker Hughes case, data collection was complete in nine weeks, after which the team turned to stakeholder expectations by role and to the decisions the assessment needed to support. Notably, the project's lens was not automation but priorities: as Caekebeke explained, "We have a strong sustainability team, and I had enough humans and employees." The point of AI at this stage is coverage and triage, not headcount reduction.

Stage 2: Establishing Industry Benchmarks and Context

Raw analysis without a reference standard produces noise. Determining which issues matter most requires an industry-specific benchmark, and the Kyushu University team turned to the Sustainability Accounting Standards Board (SASB), which identifies financially material sustainability issues by industry. The benchmark does the definitional work: it tells the model what "material" means for, say, an energy company versus a software firm, so the analysis measures disclosure against the issues that actually affect enterprise value in that sector.

For executives, the decision point in this stage is benchmark selection. The choice should reflect which stakeholders the assessment serves—investors focused on financial materiality will gravitate toward SASB-style standards, while broader stakeholder assessments may layer in additional frameworks. Whatever the choice, the benchmark must be fixed before analysis begins, because it becomes the yardstick against which every disclosure is scored.

Academic decision-support research points to a related principle: ideal benchmarks act as reference points when assessing alternatives. In multi-criteria evaluation methods used for AI-enabled ESG strategy assessment, researchers define positive and negative ideal solutions for each criterion and measure how far each alternative sits from them. The same logic applies here—benchmarks give the AI both a target and a floor, making the resulting scores interpretable rather than merely descriptive.

Stage 3: AI-Powered Semantic Analysis for Deeper Insights

This is where AI distinguishes itself from keyword counting. The Kyushu University model rates the semantic relevance of disclosures: a statement giving the exact tons of emissions a company cut scores higher than a general pledge to work toward carbon neutrality. Those scores let the researchers systematically compare, by company, year, and ESG issue, how closely disclosures address the issues SASB identifies as financially material.

The practical implication for leadership is that the analysis distinguishes substance from vagueness. A company that mentions "climate" forty times with no quantified commitments should not outrank a competitor that reports specific reductions once, with numbers. Semantic scoring surfaces that difference, which mention-counting and sentiment scoring systematically miss.

Reproducibility is a further advantage worth demanding from any implementation. Hidemichi Fujii, a professor at Kyushu University's Faculty of Economics who supervised the study, notes that the model uses zero-shot analysis: "It doesn't require researchers to manually write prompts or provide subjective guidance, making the analysis more standardized and reproducible." For a board, that matters because it reduces the risk that the assessment reflects the analyst's framing rather than the disclosures themselves.

Stage 4: Stakeholder-Specific Prioritization and Strategic Integration

The final stage converts scores into strategy, and it is deliberately human. Materiality looks different to different stakeholders—customers, investors, regulators, and employees weight issues differently—so the AI output must be read through the lens of who the decision serves and what decision it needs to support. In the Baker Hughes project, this meant mapping stakeholder expectations by role after the data collection phase, not before.

Executives should also treat AI governance as part of the ESG question itself. Research on corporate AI disclosure has found that high-density AI governance information tends to appear in reports of moderate length, while over-lengthy reports show a notable lack of materiality on key AI issues—evidence that, absent standardized reporting frameworks, companies risk using "redundant narratives" to mask a "governance vacuum." A strong governance process for the models used in the assessment is therefore integral to the integrity of the sustainability disclosures that follow.

Academic work on AI-enabled ESG strategy evaluation adds a caution: matching AI outputs with changing ESG regulations and stakeholder expectations adds a level of complication, and researchers have proposed hybrid data governance models, explainable AI, stakeholder training, and staged digital transformation as responses. The executive takeaway is that the framework is iterative—benchmarks shift, regulations evolve, and the assessment should be rerun on a defined cadence.

AI-Powered ESG Materiality Assessment: A Four-Stage Framework

The table below consolidates the four stages into a decision matrix a leadership team can use to oversee implementation. Each row pairs the executive's key action with what the AI contributes and the strategic outcome to expect.

Read across the rows, the matrix makes one dependency explicit: each stage's value depends on the one before it. Weighted inputs make benchmarking meaningful; benchmarks make semantic scores interpretable; and scores only become strategy through deliberate human prioritization.

Actionable Insights: Prioritizing ESG for Strategic Advantage

C-suite executives should initiate a pilot program for AI-powered ESG materiality assessment, focusing on integrating diverse data sources and leveraging semantic analysis against industry benchmarks. A pilot bounded to one business unit and one benchmark keeps the effort manageable while testing the full four-stage chain.

The success indicator to watch is concrete: the organization observes a measurable increase in the specificity and relevance of identified material ESG issues, leading to more targeted strategic initiatives and improved stakeholder engagement reports. If, after a defined cycle, the assessment still surfaces vague pledges rather than quantified commitments—or if source weighting and benchmark selection were skipped—the pilot has not yet delivered its value, and the gap should be closed before scaling.

Sources

• 94% of CFOs Feel ESG Pressure. AI Can Help — C3
• AI helps reveal what sustainability issues companies emphasize — Phys.org
• How Baker Hughes used AI, LLMs for ESG materiality assessments — Constellationr
• A decision-support framework for evaluating AI-enabled ESG strategies in the context of… — PMC
• MDPI