For decades, the business conversation around climate change has been focused on how to manage, mitigate, and withstand the risks and downsides. But what if there’s a more important story about opportunity that we’ve been missing? In the new article, “Tracking Business Opportunities for Climate Solutions Using AI in Regulated Accounting Reports,” published in Nature Communications, a team of co-authors including Shirley Lu, George Serafeim, and Simon Xu of the Climate and Sustainability Impact Lab at the Digital Data Design Institute at Harvard (D^3), and Mark Antonio Awada turned a large language model loose on nearly 40,000 regulatory filings from over 4,000 US firms in 47 GICS industries between 2005 and 2022. By fine-tuning a GPT to identify climate solutions, they’ve created a systematic measure of climate business opportunities across the U.S. economy.1 They found that 45% of firms in 2022 mention climate solutions in their core business descriptions, up from 20% in 2005. More importantly, the AI-powered approach reveals patterns difficult to discover using traditional analysis, from which technologies are actually driving growth to how seemingly unrelated industries are quietly converging around shared clean-tech innovations.
Key Insight: Teaching AI to Distinguish Opportunity from Noise
“We fine-tune the GPT model to perform the specific task of identifying sentences that relate to climate solutions.” [1]
The paper tackles a basic but stubborn problem: while climate risks are increasingly measured (emissions, carbon pricing exposure), there is no standardized way to track climate opportunities, the products and services that help others decarbonize. Companies don’t have to disclose climate solution revenues, and when they do, it’s often in voluntary, non-standardized formats. To fill this gap, the team fine-tuned a GPT model on 3,508 carefully labeled sentences to distinguish genuine climate solutions from generic climate-related discussion. For example, a company that manufactures electric vehicles is pursuing climate solutions; a company that uses electric vehicles in its corporate fleet is not. Then they turned the GPT on the business descriptions within 10-K filings. This is critical because it describes what a company actually sells, and unlike a press release, it is scrutinized by auditors and carries legal liability for misrepresentation.
Key Insight: AI Reveals Which Climate Technologies Actually Drive Growth
“Overall, the climate solutions measure is positively and statistically significantly associated with revenue growth.” [2]
The AI’s ability to categorize sentences into specific technology topics enabled granular analysis. And once you can measure climate solutions systematically, you can ask questions impossible to answer before, such as, ‘Do these opportunities actually translate to business performance?’ AI-derived metrics revealed that firms with a one standard deviation higher climate solutions measure experienced 2% higher revenue growth. However, the positive association is statistically significant primarily in industries where innovation is protected by patents. This suggests that commercializing climate solutions is not just about having a good idea, it’s about possessing the intellectual property to defend it. Furthermore, the AI model allowed the researchers to categorize the type of technology being discussed. They found that revenue growth was stronger for technologies with high abatement potential, solutions that can significantly reduce emissions, indicating that the market is effectively rewarding technologies that solve the biggest chunks of the climate problem.
Key Insight: The Great Convergence
“We observe a blurring of industry boundaries, with previously unrelated industries engaging in similar products because of climate solutions.” [3]
Perhaps the most striking discovery emerged when researchers visualized the AI-generated embeddings, representations of how similar different climate solution sentences are to each other. By plotting these embeddings, the researchers could analyze not just how much individual firms disclose about climate solutions, but how entire industries cluster around shared technology narratives. For example, electric vehicle topics clustered around both automobiles and capital goods (equipment manufacturing), revealing the emerging battery and electric powertrain value chain. This is more than just linguistic similarity: the study found that industry pairs with higher topic similarity exhibited higher stock return synchronicity. Essentially, if you talk about the same climate tech, your stock prices start moving together because your economic fundamentals are aligning.
Why This Matters
For business leaders, investors, and strategists, this research underscores how AI can change how we understand markets. By converting dense, regulated text into structured intelligence, AI can reveal patterns that are otherwise invisible, whether related to climate technologies, shifting consumer needs, emerging value chains, or entirely different domains. Even companies that operate far from climate-related sectors can benefit from the underlying method: using AI to mine filings, contracts, product documentation, customer feedback, and market reports for indicators of strategic change and future direction. Ultimately, this research demonstrates the potential for AI to handle sophisticated, complex analysis, and for organizations to evaluate whether their own narrative matches their strategy and competitive aspirations.
Footnote
1. The data generated from the study is publicly available at https://github.com/Climate-Solutions-Project/climate-solutions-10k.
References
[1] Shirley Lu et al., “Tracking opportunities for climate solutions using AI in regulated accounting reports,” Nat Commun 16, 9769 (2025): 2. DOI: https://doi.org/10.1038/s41467-025-64723-1
[2] Lu et al., “Tracking opportunities for climate solutions using AI in regulated accounting reports,” 6.
[3] Lu et al., “Tracking opportunities for climate solutions using AI in regulated accounting reports,” 12.
Meet the Authors
Shirley Lu is Assistant Professor of Business Administration in the Accounting and Management Unit at Harvard Business School. She is faculty within the Climate and Sustainability Impact Lab at D^3.
George Serafeim is Charles M. Williams Professor of Business Administration at Harvard Business School. He co-leads the Climate and Sustainability Impact Lab at D^3.
Simon Xu is a Postdoctoral Fellow at the Climate and Sustainability Impact Lab at D^3.
Marc Antonio Awada is Chief Innovation and Digital Strategy Officer at Brown Capital Management, Senior Lecturer at Harvard Extension School, and was formerly Head of Research and Data Science at D^3.