Is AI the GPS guiding green investment, unlocking 40 trillion dollars in ESG?

 

A paradox is unfolding before our eyes: while the world accelerates its push for sustainable investing, ESG assets in the United States, the global financial hub, have dropped from $17 trillion to $8.4 trillion in just two years. This is not due to capital flight, but rather the result of increased scrutiny, demanding greater transparency and higher standards than ever before. Globally, ESG assets are projected to surpass $40 trillion by 2030.

Could AI be the key to helping businesses stay ahead in this race, enhance their appeal to investors, attract new capital, and strengthen their brand value?

1. Green Investment Market Context

1.1 ESG Market Context

The global sustainable investment market is undergoing a significant transformation. ESG investing focused on environmental, social and governance factors is no longer merely an ethical choice, but increasingly a mainstream strategy. It integrates non-financial considerations into decision-making processes with the aim of achieving long-term and sustainable growth.

According to statistics, as of early 2022, the total value of ESG-aligned assets globally reached approximately USD 30.3 trillion, accounting for nearly 25% of total assets under management across the five largest markets: the United States, the European Union, the United Kingdom, Canada and Japan.

However, the ESG investment market is gradually entering a more mature phase. Regulators and investors are tightening standards, definitions, and evaluation methodologies in response to the rise of greenwashing [1] [2] [3].

One country that exemplifies this trend is the United States. The total value of assets classified as ESG investments fell sharply from USD 17 trillion in 2020 to USD 8.4 trillion in 2022. This decline, however, does not reflect an exodus of capital from ESG. Instead, it is largely the result of stricter assessment criteria that have rendered many investments ineligible for ESG classification [2].

 

Climate Report: North America and Europe 2023 – 2024 (Source: Internet

Across Europe, the trend towards tightening ESG standards is also taking shape. Many investments previously classified as ESG-compliant have been reclassified, as they no longer meet the new, more rigorous criteria. This reflects a broader shift from lenient ESG labelling to a stronger emphasis on transparency, accountability, and the authenticity of investment strategies [1].

Despite undergoing a period of market cleansing, the long-term outlook for ESG investing remains highly positive. Forecasts suggest that by 2030, global ESG assets could surpass USD 40 trillion [1].

In Asia, capital flows into sustainable investment activities are growing rapidly. As of Q1 2023, the total assets of sustainable investment funds in Asia (excluding Japan) reached USD 63 billion, with Taiwan leading in terms of growth rate. The ASEAN region also stands out, with a cumulative green finance market value of USD 128.8 billion by the end of 2022. Within that, sustainable bond issuance increased by 40% in 2023 compared to the previous year.

In Vietnam, the sustainable finance market reached nearly USD 900 million in 2022 (a 40% decline compared to 2021), although cumulative market size remained at around USD 2.4 billion. International green bonds and green loans accounted for approximately USD 1.5 billion, placing Vietnam second in ASEAN for green debt issuance, just behind Singapore [4].

Green Credit Outstanding 2016 – 2022 (Source: Internet

A 2023 report by PwC Vietnam revealed that 93% of Vietnamese consumers are willing to pay a premium for products made from recycled or sustainable materials, signalling a clear shift in consumer behaviour and growing awareness of sustainable development [5].

However, ESG investing continues to face significant barriers. One of the most pressing challenges is the lack of standardisation in ESG measurement and reporting on a global scale. ESG information is often fragmented, inconsistent, and difficult to access particularly for small and medium-sized enterprises or organisations in developing countries.

In addition, the low correlation between ESG ratings from different providers highlights ongoing disagreements in approaches and assessment methodologies. Furthermore, the difficulty in establishing a clear causal relationship between ESG factors and financial performance remains a subject of debate. This makes it challenging for investors to accurately quantify the impact of ESG strategies on investment returns.

1.2 Artificial Intelligence in ESG Investing

In this context, artificial intelligence (AI) is emerging as a powerful transformative tool with the potential to address many of the inherent weaknesses in ESG investing.

AI enables the automation of ESG data collection, processing and analysis at scale, drawing from a wide range of sources including financial reports, social media, news articles and sensor data from IoT devices [9].

Notably, natural language processing (NLP) allows AI to extract qualitative insights from textual data such as public sentiment, trending topics, or ESG-related events in real time. This provides a more comprehensive, dynamic and continuously updated view of a company's ESG performance [8].
 

Performance of ESG Portfolio Based on NLP (Source: Internet)


In addition, machine learning (ML) and deep learning (DL) algorithms have the capability to predict ESG scores, assess risks, and optimize investment portfolios based on multiple objectives such as returns, financial risk, and sustainability impact. AI models can even personalize investment recommendations to align with each investor’s individual values and preferences through robo-advisory platforms. Moreover, the reinforcement learning capability of AI allows for dynamic asset allocation, continuously adapting to real-time data [7].
 

Architecture for Predicting the Financial Performance of Listed Companies (Source: Internet)

The key benefits that AI brings to ESG investing include:

Enhanced efficiency and scalability: Automating data analysis processes saves time and costs while increasing the ability to handle large-scale datasets.

Real-time information updates: AI enables rapid analysis of market sentiment and risk forecasting, helping investors make more proactive decisions.

Improved risk management: AI helps identify emerging ESG risks early, such as climate change, regulatory shifts, or reputational crises.

Greater transparency and personalization: AI supports the construction of tailored “green” portfolios that reflect both financial goals and individual ethical values.

The value of AI in ESG investing goes beyond a single function, it delivers comprehensive improvements throughout the entire investment lifecycle.

2. The Personalization Mechanism of AI in ESG Stock Recommendations

For artificial intelligence to effectively learn how to process, analyze, and deliver accurate recommendations, a prerequisite is access to a vast, clean, and high-quality data source. In particular, the comprehensive integration of three key components ESG data, financial data, and investor risk appetite is crucial. Only with this foundation can AI generate comprehensive, balanced, and tailored ESG stock recommendations that suit each individual investor efficiently.

2.1 Aggregation of ESG Data

Aggregating Environmental, Social, and Governance (ESG) data from multiple sources is an essential foundation for building AI-based stock recommendation systems. This process includes the collection, processing, and standardization of information from ESG rating providers, corporate sustainability reports, and alternative data sources such as media coverage and geospatial data. The objective is to create a consistent, high-quality, and information-rich ESG data ecosystem to support sustainable investment risk assessment and opportunity identification.

Reputable ESG rating agencies such as MSCI, Sustainalytics, and CDP play a vital role in delivering in-depth insights for responsible investing. For example, MSCI employs an ESG evaluation methodology based on a company’s exposure to long-term financially relevant ESG risks, assigning scores from AAA to CCC across the Environment, Social, and Governance pillars. Similarly, Sustainalytics adopts an industry-specific ESG risk analysis approach [9].

However, the absence of unified reporting standards renders ESG data fragmented and inconsistent. A 2019 study found that MSCI’s ESG scores had a correlation coefficient of only around 0.5 with those from other rating agencies, highlighting the subjectivity and poor comparability across different assessment frameworks [10].

Traditional ESG Rating Agencies vs. AI-Driven ESG Rating Agencies (Source: Internet)

AI offers a viable solution to these challenges. By leveraging machine learning, deep learning, and natural language processing (NLP) techniques, AI can integrate diverse data sources, process unstructured information, and refresh ESG ratings on a daily basis [10]. This helps reduce human bias, improve the accuracy of ESG information, and ensure it remains up to date. AI is capable of aggregating data from corporate reports, media coverage, and alternative data sources to provide a more comprehensive and objective view of a company's ESG performance [10]. 

In addition to data aligned with international standards, internal company information and stakeholder data particularly Scope 3 emissions data play a critical role in ESG evaluation.

Scope 3 includes indirect emissions across the entire value chain, from upstream inputs and transportation to product use and end-of-life processing. In practice, Scope 3 emissions often account for 65 to 90 percent of a company's total emissions, underscoring the importance of transparency and integration of this data in sustainability assessments [11] [12].

When Scope 3 emissions are included in investment analysis, the total carbon footprint of a portfolio can increase by up to four times compared to analyses based solely on direct emissions (Scope 1 and Scope 2). This highlights a significantly higher level of environmental risk than is often apparent on the surface, especially in sectors with complex supply chains such as manufacturing, transportation, and consumer goods.

GHG Protocol Emission Scopes (Source: Internet)

Artificial intelligence (AI) can help address the challenges of evaluating Scope 3 emissions by automating the processes of data collection, standardization, and analysis. In particular, AI can be combined with geospatial analytics to identify major emission sources across a company's entire value chain.

As a result, AI not only enhances access to and processing of ESG data, but also transforms complex datasets into actionable investment insights. This enables investors to restructure their portfolios to better align with emission reduction goals and global climate commitments  [13] [14] [15].

Another ESG factor receiving growing attention is board diversity, particularly gender diversity. Numerous studies have shown that higher female representation on boards is positively associated with financial performance, especially among profitable firms  [15].

Notably, AI models such as artificial neural networks (ANNs) have been applied to analyze the nonlinear relationship between gender diversity and financial indicators such as return on assets (ROA). Unlike traditional linear models, AI has the ability to uncover complex relationships between non-financial attributes such as board composition and business performance. Consequently, gender diversity metrics go beyond ESG compliance. When effectively processed by AI, they can serve as forward-looking financial predictors, helping investors assess the quality of corporate governance in a more strategic and comprehensive manner [16] [15].<
Data Source / Type of ESG Data Main Indicators / Examples AI Integration Methods Purpose / Benefits in AI Analysis
ESG Ratings (MSCI, Sustainalytics, CDP) ESG Scores (AAA–CCC), industry rankings Machine learning, deep learning, NLP, neural networks Address data inconsistency, integrate multiple sources
Scope 3 Emissions Data Indirect emission categories, carbon intensity indices Automation, geospatial analysis, smart gap filling Standardize and verify data, climate risk assessment, portfolio adjustment
Board Diversity Gender ratio, proportion of business members Artificial Neural Networks (ANNs) Discover nonlinear relationships, predict ROA, assess corporate governance
Sustainability Reports & Media News Greenwashing indices, ESG sentiment, related controversies NLP, contextual sentiment analysis Detect ESG risks, guide employee satisfaction, handle unstructured data
Alternative Data (e.g., satellite imagery) Environmental data, plant emissions, deforestation monitoring Geospatial analysis, multimodal deep learning Independently verify sustainability claims, supply chain monitoring, emission checks

ESG Data Table and AI Integration Methods

2.2 Integrating Financial Data through AI

While ESG considerations are important, the ultimate goal of investing remains financial return. Financial data such as price-to-earnings (P/E) ratios, beta coefficients, expected returns, and revenue forecasts provides a clear picture of a company's financial health, profitability potential, and exposure to market risk.

AI enhances the accuracy of these financial forecasts and helps uncover complex relationships that traditional methods may overlook. By processing large volumes of structured and unstructured financial information, AI models can identify subtle patterns, correlations, and anomalies, offering deeper insights into a company’s financial outlook and investment potential.

AI Applications in Financial Services (Source: KPMG)

The Price-to-Earnings (P/E) ratio is a key indicator for evaluating stock value and predicting market trends. Machine Learning (ML) models can uncover non-linear relationships between earnings and a wide range of financial factors, thereby enhancing the accuracy of price estimation. A study utilizing over 5,000 input variables, combining factors from both simple and multivariate regressions, demonstrated that ML models could reduce error by up to 68% compared to simple regression and 26% compared to traditional multivariate regression models [17] [18].

Meanwhile, beta measures systematic risk and the relationship between stock return volatility and the market. Traditional methods often provide static and unstable beta estimates. NeuralBeta, an advanced deep learning approach, delivers dynamic and more accurate beta estimates by capturing non-linear and time-varying relationships, thereby improving portfolio optimization effectiveness [19].

Expected return is a fundamental concept in asset valuation and portfolio management. In recent years, ML models have demonstrated superior performance in predicting financial returns thanks to their ability to extract complex signals from multiple variables such as the Earnings-to-Price (EP) ratio, capital expenditure growth (CAPEX), and income returns. Compared to traditional regression models, ML-based forecasts are significantly more accurate, particularly over medium- to long-term horizons [17] [18].

In addition, AI is increasingly applied in revenue forecasting, a critical element in corporate financial analysis. AI models can analyze large volumes of historical data, automatically standardize information, adjust assumptions in real-time, and forecast revenues with up to 92% accuracy at the SKU (Stock Keeping Unit) level . This not only enables companies to gain deeper insights into growth drivers but also supports agile strategic decision-making, thereby positively influencing expected return forecasts and stock valuation models  [21].

In summary, the core strength of AI in finance lies in its ability to process and analyze the inherent non-linear and dynamic relationships within financial data. When fundamental indicators such as P/E, beta, and expected return are effectively integrated into sophisticated machine learning algorithms, predictions and risk assessments become significantly more accurate and practical. This contributes to building superior and sustainable ESG investment portfolios in the face of continuously volatile markets  [22] [23] [24].

Financial Indicator Role in Traditional Finance AI/ML Integration Benefits of AI/ML Integration
P/E Ratio Used for stock valuation and determining a company’s intrinsic value Used as an input feature in machine learning and deep learning models such as neural networks (ANNs), XGBoost Helps AI identify complex relationships between factors, accurately predict stock prices and volatility
Beta Measures the risk and volatility of a stock compared to the market Estimated using deep learning methods such as NeuraBeta (Deep neural networks) Provides more accurate beta estimates, captures time-varying changes, enhances risk management
Expected Return Serves as a basis for making investment decisions Forecasted using general machine learning models and predictive analytics Improves forecast accuracy, detects complex patterns to optimize portfolio allocation
Revenue Forecasting Supports business planning and forecasting Applies machine learning, time series forecasting, regression algorithms, and scenario analysis Delivers more accurate revenue forecasts, enables real-time analysis, and supports strategic decision-making

Table of Financial Indicator Integration into AI/ML Models

2.3 Risk Appetite

Each investor has a different level of risk tolerance. Integrating risk appetite, measured through surveys or historical investment behavior, is essential to ensure that AI-generated recommendations are not only sustainable and potentially profitable but also aligned with each individual's risk-bearing capacity.

Artificial Intelligence (AI) can analyze actual investment behavior to construct more accurate risk profiles, and even detect and mitigate behavioral biases a factor often overlooked in traditional models.

At the organizational level, overall risk appetite is shaped by the degree of risk an organization is willing to accept in order to achieve its strategic objectives. In AI-driven financial advisory contexts, evaluating an individual investor’s risk tolerance typically relies on various methods most commonly structured questionnaires using Likert scales, which allow investors to self-report their preferences and risk acceptance levels [26].

However, recent studies indicate that investors with higher risk tolerance often exhibit a more positive perception of behavioral control, making them more confident when interacting with AI platforms. They are more open to experimenting with algorithmic decisions and readily integrate AI-generated information into their financial strategies  [26]. 

Conversely, risk-averse investors or those with low risk appetite tend to develop a cautious attitude toward AI-based financial advisory tools. This is often due to perceptions of opacity, lack of flexibility, or concerns that algorithms cannot adequately process qualitative factors such as market sentiment or policy impacts.

Risk Tolerance Scale (Source: Internet)

Therefore, the integration of historical investment behavior analysis using Machine Learning (ML) has opened a new, more accurate approach to constructing investor risk profiles. Unlike traditional surveys which are often influenced by cognitive biases,  ML leverages real-world data such as stock trading history, credit activity, social media behavior, and simulated trading tests to infer an investor’s true risk tolerance. From this, AI can:

  • Objectively infer risk appetite based on observable behavior
     
  • Cluster investors with similar risk profiles using clustering techniques
     
  • Detect potential behavioral biases by analyzing transaction frequency, holding periods, and reactions to market volatility

Through these capabilities, AI not only enables the development of personalized investment strategies, but also ensures alignment with each investor’s risk appetite and long-term financial goals particularly critical in the context of growing interest in ESG investing [26] [27].

Going further, modern AI technologies have increasingly integrated behavioral finance factors into risk analysis models, enabling more comprehensive and adaptive risk assessment frameworks.

Going further, modern AI technologies have deeply integrated behavioral finance factors into risk analysis models. Behavioral finance suggests that emotions and cognitive biases such as loss aversion, overconfidence, or herding behavior significantly influence investment decisions, often leading to irrational or anomalous market behavior [28]. AI is capable of continuously monitoring behavioral traits such as trading frequency, holding periods, reactions to price fluctuations, and consistency in portfolio rebalancing strategies, thereby identifying key biases, including:
  • Loss aversion: Investors hold onto losing assets for too long instead of cutting losses in a timely manner [28]
     
  • Overconfidence: Engaging in trades that exceed personal or market risk thresholds [28]
     
  • Tax strategy biases: Accidentally triggering disallowed transactions such as wash sales  [28]

Beyond monitoring, AI can generate realistic market simulations to allow investors to observe their own behavioral responses, recognize biases, and adjust accordingly. When behavioral risks are detected, AI can provide real-time feedback, offering suggestions such as portfolio rebalancing, re-evaluating investment decisions, or adjusting risk exposure [29].

This capability is particularly valuable in ESG investing, where personal emotions and ethical values often play a significant role in decision-making. AI plays a critical role in enhancing discipline and objectivity. By delivering instant bias alerts and supporting data-driven decision-making, AI helps investors stay aligned with their long-term sustainability goals while improving the efficiency and resilience of their investment strategies [28].

 

3. How Companies Apply AI in Investment Decisions

The application of AI in ESG investment decisions is no longer just theoretical it is being concretely implemented with real-world results by companies like Clarity AI and GreenFi.

3.1 Clarity AI Enables NLP-Based Analysis of Over 100,000 Articles Daily

Clarity AI has rapidly deployed advanced machine learning (ML) and natural language processing (NLP) models to analyze vast volumes of fragmented ESG data, transforming them into transparent and actionable insights.

The platform can process over 100,000 news articles per day, extracting real-time information from unstructured data sources. The integration of proprietary NLP technology with AI allows Clarity AI to  [30]:

  • Automatically analyze sustainability reports
     
  • Convert static disclosures into dynamic, real-time analyses
     
  • Eliminate subjective bias in ESG ratings through objective algorithms
     

One of Clarity AI’s most remarkable capabilities lies in its ability to process ESG data at a scale three times larger than its competitors. The platform covers over 70,000 companies, 400 governments, and 430,000 investment funds [30] [31]. This has enabled Clarity AI not only to enhance the speed and scale of ESG data processing, but also to improve objectivity, consistency, and data reliability, significantly reducing the risk of poor investment decisions caused by missing or biased data.

Environmental, Social, and Governance (ESG) Risk Analysis by Clarity AI (Source: Clarity AI)

Clarity AI’s operational process is built on a comprehensive framework, leveraging AI to collect, process, and transform fragmented ESG data into deep, consistent analyses. The data is continuously updated from millions of sources and standardised according to major international regulatory frameworks, such as the SFDR, CSRD, and EU Taxonomy [32].

The platform also allows for the flexible customisation of ESG indicators to align with each investor’s unique strategy and investment objectives.

Following the standardisation phase, Clarity AI assesses ESG risks both at the portfolio level and for individual assets, while providing tools for portfolio rebalancing and tailored recommendations that reflect the investor’s risk appetite and financial goals.

Finally, the system automates compliance reporting processes, helping institutions save time, reduce operational costs, and enhance transparency. As a result, financial institutions can more effectively meet the growing regulatory demands associated with sustainable investing [33] [32] [34].

3.2 GreenFi: AI-Driven ESG Due Diligence and Greenwashing Prevention

While Clarity AI focuses on portfolio optimisation through artificial intelligence, GreenFi is a specialised SaaS platform designed for ESG due diligence and climate risk management. The platform integrates sustainability data from a wide range of sources including corporate disclosures, industry databases, and public information and combines this with auditing tools and AI-powered predictive recommendations to deliver a comprehensive ESG assessment.

A key distinguishing feature of GreenFi is its use of explainable AI (XAI), which ensures that every decision generated by the system is transparent, comprehensible, and auditable in plain language [35]. GreenFi leverages natural language processing (NLP) to extract ESG signals from unstructured data, including ESG reports, news articles, and legal documents, enabling the independent quantification of ESG performance  [36]. 

Crucially, GreenFi's AI model extends beyond conventional machine learning techniques by incorporating human-level causal reasoning. This approach surpasses basic statistical correlations to deliver more accurate and actionable predictive insights  [35][36]. 

The platform is capable of processing billions of data records and automating millions of ESG-related decisions daily, supporting various financial instruments such as green bonds, trade finance, sustainability-linked loans (SLLs), and sustainability-linked bonds (SLBs) [36].

 

ESG Data Analysis and Management System by GreenFi (Source: GreenFi

In addition to its analytical capabilities, GreenFi offers a comprehensive 360-degree view of ESG risks at the levels of client, asset, portfolio, and supplier [37]. The platform enables institutions to identify, assess, and mitigate risks through real-time monitoring, risk scoring systems, and scenario analysis [37].

GreenFi’s ESG risk library spans 17 sectors and over 135 sub-sectors, encompassing more than 36,000 distinct risk types organised in a granular taxonomy. The platform also integrates a decarbonisation and mitigation initiative library, featuring more than 3.6 million science-based actions [37]. This empowers institutions to manage ESG risks not only within investment operations but also across the entire enterprise ecosystem including supply chains, lending portfolios, and insurance underwriting [37]. 

GreenFi’s decision-support capabilities are particularly notable. The platform helps financial institutions and corporations align capital allocation with strict ESG criteria, ensuring investments are directed toward truly sustainable projects and reducing regulatory compliance risks. The AI engine provides real-time ESG assessments, automated compliance checks, and plays a pivotal role in greenwashing prevention by ensuring that capital is transparently and efficiently allocated.

Instead of relying on complex scoring models, GreenFi adopts a binary evaluation framework “Good or Bad”, “Yes or No” to simplify decision-making and enhance compliance adherence [37]. 

Moreover, GreenFi automates the entire ESG data collection and reporting process, in line with major international standards such as GRI, SASB, and TCFD. The platform continuously updates regulatory changes, monitors compliance gaps, and maintains audit-ready documentation. It is certified with high-level security and reliability credentials, including Vulnerability Assessment and Penetration Testing (VAPT) and SOC-2, and operates with an impressive 99.98% uptime.

This high level of automation minimises manual intervention, reduces human error and regulatory breach costs, and allows organisations to focus on core strategy rather than time-consuming operational tasks [38] [39].

4. Challenges and Risks in Applying AI to ESG Stock Recommendations

While artificial intelligence is increasingly demonstrating clear benefits, it is not a silver bullet and comes with significant challenges.

Key risks include poor-quality and non-standardized ESG data, algorithmic bias, lack of transparency and explainability in models (the “black box” issue), phenomena such as “AI-washing” and “greenwashing,” operational risks amid volatile markets, and the environmental impact of running AI systems themselves.

If left unaddressed, these issues can lead to misguided investment decisions, erode trust in sustainable finance, and result in negative social and environmental consequences.

4.1 ESG Data Quality

A major barrier to the application of AI in ESG investing is the lack of data standardisation, stemming from the coexistence of multiple reporting frameworks such as GRI, SASB, TCFD, CSRD, and ISSB. Each framework offers distinct definitions, metrics, and methodologies, making cross-comparison difficult and resulting in inconsistent data sets [40][41]. According to Diligent, over 60% of companies face significant challenges in collecting and reporting ESG data  [42]. 

The data are often fragmented across various sources, including internal reports, third-party databases, and unstructured data from news articles and social media platforms [43][44]. While AI excels at processing large-scale data, the tasks of data cleaning, integration, and harmonisation remain time-intensive, and poor data quality can undermine the reliability and credibility of AI-driven investment recommendations [45][46].

Correlation Between ESG Rating Providers (Source: Internet)

Moreover, only around 30% of small and medium-sized enterprises (SMEs) have access to ESG standards due to resource constraints and the lack of standardised metrics [47]. As a result, AI models tend to be biased towards large corporations, which leads to scale-related skew and the systematic exclusion of high-potential SMEs from ESG assessments. This limits their ability to access green financing and participate in sustainable capital flows  [48][49]. 

In addition, many ESG reports contain unstructured data, such as free-text disclosures, satellite imagery, or content from social media [50][51]. While Natural Language Processing (NLP) and Machine Learning (ML) can handle such data types, the processes of standardisation, verification, and auditability remain complex and resource-intensive [52].. ESG scoring also involves a significant degree of subjectivity, which results in substantial variations across rating providers and raises the risk that AI models may amplify existing biases rather than correct them  [54][52][53].

4.2 Algorithmic Bias and Fairness Risks

AI systems are only as trustworthy as the data used to train them. However, much of the historical data available for training reflects longstanding societal inequalities across various domains such as recruitment, credit scoring, and healthcare [55][56]. For instance, a clinical algorithm used in the United States underestimated the healthcare needs of Black patients because it relied on healthcare spending as a proxy for actual medical need [57]. Similarly, tools such as facial recognition software, credit scoring models, and automated recruitment systems have demonstrated gender and racial biases in their outputs  [58]. 

Within the ESG context, if AI models assess social or governance factors based on biased datasets, the resulting investment recommendations may lack fairness, thereby undermining the goals of social equity and inclusive finance  [59].

Bias can also emerge through the weighting of ESG indicators, where subjective value judgements or design biases in algorithm construction influence outcomes. The choice of AI models further contributes to this issue, as different models are optimized for different types of data, potentially leading to model selection bias [60]. Furthermore, current ESG data is predominantly sourced from large corporations in developed markets, which skews ESG scores and increases the risk of overlooking high-potential small firms in emerging markets [61].

ESG Assessment Process from Data to Final Outcome (Source: Internet)

As a result, sustainable finance algorithms may make unfair capital allocation decisions, exacerbating existing inequalities [59]. For example, if algorithms consistently prioritize green projects in developed regions, opportunities and data will become concentrated on one side, while projects in developing areas may be undervalued or excluded due to a lack of data [58]. This not only leads to higher borrowing costs or denial of green financing but also hinders global efforts toward sustainable development 57][58]. 

4.3 Governance, Legal, and Operational Challenges

Beyond the inconsistency of measurement systems and the persistence of unclean green data, the issue of the “black box,” especially with deep learning models, remains a major barrier to applying AI in ESG  [61].

In the ESG field, Explainable AI (XAI) plays a crucial role in helping users understand the factors influencing evaluation outcomes, such as labor disputes or low social audit scores. A lack of transparency in AI decision-making not only undermines trust but also complicates result verification and regulatory compliance  [62][63].

The rising phenomena of AI overstatement (“AI-washing”) and fabricated sustainability efforts (“greenwashing”) involve the misuse of AI to selectively highlight positive data while hiding negative information, misleading stakeholders about a company’s sustainability level. Such behavior can lead to serious legal consequences as authorities like the SEC begin to address false claims [57][58]. Additionally, issues of privacy, accountability, and the carbon footprint generated by AI operations pose significant ethical challenges  [62][58].

AI may perform poorly when market conditions change suddenly due to overreliance on historical training data, leading to risks of overfitting and instability in real-world environments [63]. Although AI can help detect risks in real time  [64], many organizations adopt model testing with unseen data and limit model complexity to enhance generalization. However, excessive dependence on AI and lack of human oversight increase the risk of severe operational failures.

 


Infographic: 4 Steps in the Human-in-the-Loop Process (Source: Internet)

Combining Explainable AI (XAI) with Human-in-the-Loop (HITL) is an essential solution to ensure fairness, transparency, and accountability in applying AI to ESG. HITL allows human intervention in the decision-making process to detect biases and adjust models, while XAI clarifies the reasoning behind each outcome, enhancing auditability and user trust. This approach is not only a technical solution but also an ethical requirement in a sensitive field like ESG  [65][66].

-------

Dung Tran

BambuUP owns a strong resource network, ready to accompany businesses. We have many years of experience in Open Innovation and connections with top industry experts, helping businesses explore sustainable solutions and create long-term value.

We have collaborated with many major companies such as EVN, Heineken Vietnam, FASLINK, DKSH Smollan... in publishing open innovation challenges.

BambuUP proudly stands as a trusted strategic partner, always supporting businesses in innovation activities and a strong green transformation process.

To not miss the latest weekly news about Green Innovation in Vietnam, you can:

Copied link

Share:

Related News

Bìa 1 (1).jpg
Sat, 07/19/2025

Động cơ nào giúp “hạ nhiệt” cho trái đất? Đọc ngày 4 giải pháp lật ngược cục diện biến đổi khí hậu!

Để hạn chế sự nóng lên toàn cầu ở mức 1.5°C ước tính rằng cần phải loại bỏ 6 gigatonne CO2 tương đương với lượng phát thải từ hơn 1.5 triệu nhà máy điện than hay khí thải từ hơn 1.3 tỷ ô tô khỏi khí quyển mỗi năm. Điều này đặt ra những thách thức vô cùng lớn cho các doanh nghiệp. Vậy làm cách nào để doanh nghiệp vừa có thể đáp ứng yêu cầu xanh mà lại không tốn kém chi phí? Đến với số Innovation of the Week tuần này, BambuUP sẽ giới thiệu 4 giải pháp nổi bật trong lĩnh vực công nghệ khí hậu
TYPES OF INNOVATION (1).png
Fri, 07/18/2025

Các loại hình đổi mới sáng tạo và những nhân tố ảnh hưởng chính

Cùng BambuUP khám phá 18 loại hình đổi mới sáng tạo phổ biến, từ đổi mới mô hình lợi nhuận, đổi mới sản phẩm và giải pháp, đổi mới trải nghiệm khách hàng, đến đổi mới gia tăng, đổi mới bền vững, và hơn thế nữa.
Thumbnail InnovationUP.png
Wed, 07/16/2025

Green Fintech: The express gateway for businesses to access sustainable finance

As ESG standards increasingly become the "entry ticket" to global markets, green finance is emerging as a major advantage for many businesses. Instead of getting stuck in legal barriers, complex procedures, and traditional credit profiles, Green FinTech is paving a new track one where businesses can accelerate access to capital, automate sustainable governance, and build transparency into every ESG data stream. The question is no longer “Should we?”, but rather: Is your business ready to get on board?