DESC:BACKGROUND
Technical Field
The embodiments herein generally relatetoEnvironmental, Social, and Governance (ESG) risk management, and more particularly, to a system and a method for generating a machine learning score using an artificial intelligence (AI) model that indicates a client’s ESG (Environmental, Social, and Governance) risk level and classify the client based on the machine learning generated score.
Description of the Related Art
In the wake of the global sustainability movement, institutions are being encouraged to consider sustainability outcomes in their lending decisions.For example, alliances like the Net Zero Banking Alliance mandate that banks need to work towards net zero emission targets by 2050. Accordingly, the banksparty to the Alliance, need to consider the ESG (Environmental, Social, and Governance) credentials of their lending portfolios to ensure that their lending decisions are having a positive impact on the planet.In short, the bank needs to evaluate the ESG risk of their clients before making lending decisions to ensure that the loans offered by them have a positive contribution towards their ESG credentials. Further, the banks need to have a clear investment strategy in place, to ensure that their lending decisions are aligned with their ESG goals. The banks need to manage the ESG risks associated with their lending portfolios to facilitate the global transition of the real economy to net-zero emissions. ESG risk management is also performed by the investors and other stakeholders involved in financial decision-making to ensure sustainable outcomes of their financial decisions.
However, evaluating ESG risk associated with a company is a very complex task and requires collecting and analyzing relevant ESG data extracted from varieddata sources such as annual reports, third-party sources, ESG rating agencies, etc. Many initiativeshave been launched recently to address concerns regarding ESG-related risk management in financial decision-making.
There exist several organizations and agencies (for example, MSCI, Sustainalytics, and RobecoSAM) that provide ESG data and ratings to help investors, stakeholders, banks, and other financial institutions to assess the ESG risks and opportunities associated with different companies. However, it is unclear what mechanism these agencies utilize to analyze and rate company ESG performance.
There are variousexisting report frameworks and standards, such as the Global Reporting Initiative (GRI), the Sustainability Accounting Standards Board (SASB), and the Task Force on Climate-related Financial Disclosures (TCFD),which aims at promoting consistent and comparable disclosure of ESG-related information. Further,a few banks in collaboration with external experts and consultants have developed their own internal ESG risk management framework and tools that involve a range of qualitative and quantitative assessments, as well as scenario analysis and stress testing, to identify and manage ESG-related risks in their lending portfolios. However, these existing solutions do not compare the ESG performance of different identified peer companies, and thus fail to provide a comprehensive understanding of the client's ESG performance. In addition, the existing solutions do not speak about automation of the data acquisition and sorting process and a comprehensive and integrated approach to estimate ESG risk and performance of a client.
Accordingly, there remains a need to address the aforementioned technical drawbacks in existing solutions toestimatethe ESG risk score of a client. Accordingly, there remains a need to provide a comprehensive and integrated approach for a method of generating machine learning scores using AI model indicating ESG risk level by the institution to ensure their lending decisions are aligned with their investment strategy and contribute towards a sustainable outcome.
SUMMARY
The primary objective of the present invention is to provide an AI-enabled platform that offers an automated, integrated, and comprehensive approach to evaluate the ESG risk score associated with a client. The present invention uses a comprehensive approach in determining ESG-risk scores associated with their lending portfolios to ensure that their lending decisions contribute towards a sustainable future.
According to the first aspect of the present invention, a system for generating a machine learning score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score is provided. The system includesa server with memorya set of instructions anda processor to execute the set of instructions. The processor is configured toobtain ESG-related data including annual reports, ESG reports, and sustainability reports of the client from a plurality of data sources and peer clients and a public sentiment data related to the client from media and public reports, generate using the AI model, a first machine learning generated score by analyzing the ESG-related data of the client and a second machine learning generated score by analyzing the public sentiment data of the client, and generate an aggregate machine learning score based on the first machine learning generated score and the second machine learning generated score for classifying the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score. The aggregate machine learning generated score is equal to the first machine learning generated score when the second machine learning generated score is greater than or equal to the first machine learning generated score and the aggregate machine learning generated score is an average of the first machine learning generated score and the second machine learning generated score when the second machine learning generated score is lesser than the first machine learning generated score.
In some embodiments, the server is communicatively connected with a first computing device associated with the client being a potential borrower and a second computing device associated with an organization involved in financial decision-making. The server is configured to communicate the machine learning-generated score to the second computing device associated with the organization.
In some embodiments, the first machine learning generated score is determined by (i) extracting and analyzing data points from the ESG-related data, (ii) assigning Key Performance Indicators (KPIs) to categories comprising environment, social and governance based on themes, (iii) assigning the categories to ESG pillars, wherein the ESG pillars comprise environmental pillar, social pillar, and governance pillar, (iv) normalizing the data points by standardizing KPI values, wherein the KPI values are standardized by dividing each KPI value by the client’s revenue for the fiscal year, (v) determining the percentile score for each data point, (vi) aggregating individual data point scores to obtain category scores, (vii) determining a weighted average for each category based on weights, wherein the weight for the environment pillar and the social pillar comprises 45% and the weight for the governance pillar comprises 10%, and (viii) determining the first machine learning generated score by summing the weighted average for each category.
In some embodiments, the second machine learning-generated score is determined by (i) performing sentiment analysis on the public sentiment data using large language models (LLMs) to classify the public sentiment data into positive, negative, and neutral categories, (ii) assigning a score of 100 to clients with no identified controversies, (iii) applying severity weights to controversy counts for addressing market cap bias, (iv) determining the count of controversies per client and multiplying the count of controversies by the severity weight based on the client’s market cap class, and (v) applying a pre-defined percentile scoring methodology to derive the second machine learning generated score.
In some embodiments, the KPIs are categorized and assigned a polarity based on their context, the polarity being either positive or negative.
In some embodiments, the processor is configured to analyze negative public sentiment data using the large language models (LLMs) to identify specified controversial topics for determining the second machine learning generated score.
In some embodiments, the severity weights have 0.33 for large-cap clients, 0.67 for mid-cap clients, and 1 for small-cap clients.
In some embodiments, the predefined percentile scoring methodology implements a plurality of factors comprising a number of peer clients with values less than the client, a number of peer clients with same value as the client, and a number of peer clients with no values for a corresponding KPI or ESG-related metric.
In some embodiments, the processor is configured to generate (i) a first survey questionnaire for the client, wherein the first survey questionnaire comprises a set of questions based on ESG categories, and (ii) a second survey questionnaire for the client, wherein the second survey questionnaire comprises a set of questions based on ESG Controversies.
In some embodiments, the ESG categories comprise one or more Business Model Resilience, Materials Sourcing & Efficiency, Physical Impacts of Climate Change, Product Design & Lifecycle Management, Supply Chain Management, Air Quality, Ecological Impacts, Energy Management, GHG Emissions, Waste & Hazardous Materials Management, Water & Wastewater Management, Employee Engagement, Diversity & Inclusion, Employee Health & Safety, Labor Practices, Business Ethics, Competitive Behavior, Critical Incident Risk Management, Management of the Legal & Regulatory Environment, Systematic Risk Management, Access and Affordability, Human Rights & Community Relations, Product Quality and Safety, Selling Practices and Customer Welfare, Data Security, or Human Rights & Community Relations.
In some embodiments, the ESG controversies comprise one or more Anti-competition controversies, Business ethics controversies, Intellectual property controversies, Critical countries controversies, Public health controversies, Tax fraud controversies, Child labour controversies, Human rights controversies, Management compensation controversies, Consumer controversies, Customer health and safety controversies, Privacy controversies, Product access controversies, Responsible marketing controversies, Responsible R&D controversies, Environmental controversies, Accounting controversies, Insider dealings controversies, Shareholder rights controversies, Diversity and opportunity controversies, Employee health and safety controversies, Wages or working condition controversies, or Strikes.
In some embodiments, the processor is configured to (i) present the first survey questionnaire and the second survey questionnaire on the first computing device associated with the client, and (ii) determine a first self-assessment score of the client based on the responses received from the client to the first survey questionnaire and a second self-assessment score based on the responses received from the client to the second survey questionnaire.
In some embodiments, the processor is configured to determine a deviation between the first self-assessment score, the second self-assessment score, and the aggregate machine learning generated score, wherein a deviation indicates a disparity between the client’s perception and the assessed ESG risk levels.
According to second aspect of the present invention, a method for generating a machine learning score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score is provided. The method includes the steps of obtaining (i) ESG-related data including annual reports, ESG reports, and sustainability reports of the client from a plurality of data sources and peer clients, and (ii) a public sentiment data related to the client from media and public reports, generating using the AI model, (i) a first machine learning score by analyzing the ESG-related data of the client, and (ii) a second machine learning score by analyzing the public sentiment data of the client, generating an aggregate machine learning generated score based on the first machine learning generated score and the second machine learning generated score and classifying the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score. The aggregate machine learning generated score is equal to the first machine learning generated score when the second machine learning generated score is greater than or equal to the first machine learning generated score and the aggregate machine learning generated score is an average of the first machine learning generated score and the second machine learning generated score when the second machine learning generated score is lesser than the first machine learning generated score. The first machine learning-generated score is determined by (i) extracting and analyzing data points from the ESG-related data, (ii) assigning KPIs to categories comprising environment, social and governance based on themes, (iii) assigning the categories to ESG pillars, wherein the ESG pillars comprise environmental pillar, social pillar, and governance pillar, (iv) normalizing the data points by standardizing KPI values, wherein the KPI values are standardized by dividing each KPI value by the client’s revenue for the fiscal year, (v) determining the percentile score for each data point, (vi) summing individual data point scores to obtain category scores, (vii) determining a weighted average for each category based on weights, wherein the weight for the environment pillar and the social pillar comprises 45% and the weight for the governance pillar comprises 10%, (viii) determining the first machine learning score by summing the weighted average for each category. The second machine learning score is determined by (i) performing sentiment analysis on the public sentiment data using large language models (LLMs) to classify the public sentiment data into positive, negative, and neutral categories, (ii) assigning a score of 100 to clients with no identified controversies, (iii) applying severity weights to controversy counts for addressing market cap bias, wherein the severity weights comprises 0.33 for large-cap clients, 0.67 for mid-cap clients and 1 for small-cap clients, (iv) determining the count of controversies per client and multiplying the count of controversies by the severity weight based on the client’s market cap class, and (v) applying a pre-defined percentile scoring methodology to derive the second machine learning score.
In some embodiments, KPIs are categorized and assigned a polarity based on their context, the polarity being either positive or negative.
In some embodiments, the method includes analyzing negative public sentiment data using the large language models (LLMs) to identify specified controversial topics for determining the second machine learning-generated score.
In some embodiments, the predefined percentile scoring methodology implements a plurality of factors comprising a number of peer clients with values than the client, a number of peer clients with same value as the client, and a number of peer clients with no values for a corresponding KPI or ESG-related metric.
In some embodiments, the method includes generating (i) a first survey questionnaire for the client, and (ii) a second survey questionnaire for the client, wherein the first survey questionnaire comprises a set of questions based on ESG categories, wherein the second survey questionnaire comprises a set of questions based on ESG Controversies.
In some embodiments, the ESG categories include Business Model Resilience, Materials Sourcing & Efficiency, Physical Impacts of Climate Change, Product Design & Lifecycle Management, Supply Chain Management, Air Quality, Ecological Impacts, Energy Management, GHG Emissions, Waste & Hazardous Materials Management, Water & Wastewater Management, Employee Engagement, Diversity & Inclusion, Employee Health & Safety, Labor Practices, Business Ethics, Competitive Behavior, Critical Incident Risk Management, Management of the Legal & Regulatory Environment, Systematic Risk Management, Access and Affordability, Human Rights & Community Relations, Product Quality and Safety, Selling Practices and Customer Welfare, Data Security, and Human Rights & Community Relations.
In some embodiments, ESG controversies include Anti-competition controversies, Business ethics controversies, Intellectual property controversies, Critical countries controversies, Public health controversies, Tax fraud controversies, Child labour controversies, Human rights controversies, Management compensation controversies, Consumer controversies, Customer health and safety controversies, Privacy controversies, Product access controversies, Responsible marketing controversies, Responsible R&D controversies, Environmental controversies, Accounting controversies, Insider dealings controversies, Shareholder rights controversies, Diversity and opportunity controversies, Employee health and safety controversies, Wages or working condition controversies, and Strikes.
In some embodiments, the method includes(i) presenting the first survey questionnaire and the second survey questionnaire to the first computing device associated with the client, and (ii) determining a first self-assessment score of the client based on the responses received from the client to the first survey questionnaire, and a second self-assessment score based on the responses received from the client to the second survey questionnaire.
In some embodiments, the method includes determining a deviation between the first self-assessment score, the second self-assessment score, and the aggregate machine learning score, wherein the deviation indicates a disparity between the client’s perception and the assessed ESG risk levels.
Thus, the present invention provides an efficient and accurate system for generating ESG risk levels for an institution. The system utilizes a self-assessment questionnaire combined with an AI-powered engine designed to extract ESG datapoints. The self-assessment questionnaire evaluates the ESG risk of potential borrowers based on their internal operations and practices. This integrated approach facilitates a thorough and expedited evaluation of ESG risks associated with borrower companies. By leveraging this system, financial institutions benefit from reduced processing time and enhanced precision in identifying high-risk borrowers, thereby optimizing decision-making processes in lending activities.The AI model extracts ESG data points from annual reports and sustainability reports of borrower companies and their peer group companies, automating the data acquisition and sorting process. This enables visualization and comparison of the ESG performance of borrowers and peers. This approach provides a benchmark for the borrower's ESG performance and enables the bank to identify outliers and better assess the borrower's ESG risks.
Further, the use of publicly available sentiment analysis that highlights key sustainability controversies associated with borrower and peer group companies enables banks or financial institutions to gain a deeper understanding of the borrower's reputation and potential risks associated with lending to them. This approach provides a more complete picture of the borrower's ESG risks and enables them to make more informed lending decisions.The combination of these elements/steps used for ESG risk score generation enables institutions to make more informed lending decisions that are technically and economically superior to existing solutions.
BRIEF DESCRIPTION OF THE DRAWINGS
The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:
FIG. 1 is a functional block diagram that illustrates a system for generating a machine learning score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score;
FIG. 2 is a block diagram that illustrates the server of FIG. 1 according to some example embodiments herein;
FIG.3 is an exemplary architectural diagram for a multi-agent workflow of the system of FIG. 1according to some example embodiments herein;
FIG. 4 is an exemplary structural diagram for a multi-prompt engine of the system of FIG.1 according to some example embodiments herein;
FIG. 5 is an exemplary structural diagram for an enablement-specific multi-agent framework of the system of FIG. 1 according to some example embodiments herein;
FIG. 6illustrates a sample interface of a company’s dashboard, showing its ESG profileaccording to some example embodiments herein;
FIG. 7 is a flow diagram that illustrates a method for generating a machine learning generated score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score according to some embodiments herein; and
FIG 8. is a schematic diagram of a computer architecture according to some embodiments herein.
DETAILED DESCRIPTION OF THE DRAWINGS
The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein.The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein.Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
Referring now to the drawings, and more particularly to FIGS. 1 through8, where similar reference characters denote corresponding features consistently throughout the figure’s, preferred embodiments are shown.
FIG. 1 is a functional block diagram that illustrates a system 100for generating a machine learning score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score. The system 100 includes a server 104 with memory a set of instructions and a processor to execute the set of instructions. The server 104 is configured to obtain ESG-related data including annual reports, ESG reports, and sustainability reports of the client from a plurality of data sources (102A, 102B, and 102N) and peer clients and a public sentiment data related to the client from media and public reports. The server 104 is configured to generate using an AI model 106, a first machine learning score by analyzing the ESG-related data of the client and a second machine learning generated score by analyzing the public sentiment data of the client, and generate an aggregate machine learning score based on the first machine learning generated score and the second machine learning generated score for classifying the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score. The aggregate machine learning generated score is equal to the first machine learning generated score when the second machine learning generated score is greater than or equal to the first machine learning generated score and the aggregate machine learning generated score is an average of the first machine learning generated score and the second machine learning generated score when the second machine learning generated score is lesser than the first machine learning generated score.
The system 100 includes a first computing device 110 associated with the client being a potential borrower and a secondcomputing device 112. The first computing device 110 and the second computing device 112 is communicatively connected with the server 104 via network 108.The computing devices 110, 112may include, but is not limited thereto, a mobile phone, a tablet, a Personal computer, a laptop, a server, automobiles, a desktop computer, or Internet of Things (IoT) devices. The computing devices 110, 112 have network 108 capabilities and include at least one processor, at least one memory, an audio-visual unit, and at least one input-output unit, all operably coupled to each other to implement functionalities known in the art. The first computing device 110 is operated by the client being the potential borrower and may include any customer, individual, or an entity/ company requiring financial assistance, whereasthe second computing device 112 is operated byan institution/ organization, for example, a bank, involved in financial decision-making.The network 108may be a combination of a wired network or a wireless network and includes any type of computer networking arrangement used to exchange data.
The second computing device 112 may be in communication with one or more ESG rating agencies to obtain third-party ESG evaluation of the client's ESG performance therefrom. In some embodiments, the AI model 106 is hosted in the second computing device 112. The system 100 enables institutions to conductan automated, efficient, and accurate comprehensive evaluation of ESG riskscores associated with their lending decisions to contribute towards a sustainable future.The AI model 106 optionally be implemented separately on a cloud computing platform or be hosted on a dedicated server machinehaving networking capabilities.
FIG. 2 is a block diagram that illustrates the server 106of FIG. 1according to some example embodiments herein. The server106 includesdatabase 212 storing a plurality of modulesand the AI model 106 to generate the machine learning score that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score. The plurality of modules includes a data obtaining module 202, a first machine learning score generating module 204, a second machine learning score generating module 206, an aggregate machine learning score generating module 208, and a classification module 210. The data obtaining module 202 is configured to (i) ESG-related data of the client from a plurality of data sources (102A, 102B, and 102N) and peer clients, wherein the ESG-related data comprises annual reports, ESG reports, and sustainability reports of the client and (ii) a public sentiment data related to the client from media and public reports. The first machine learning score generating module 204 is configured to generate the first machine learning score using the AI model 106, the second machine learning score generating module 206 is configured to generate the second machine learning score using the AI model 106 and the aggregate machine learning score generating module 208 is configured to generate the aggregate machine learning score using the AI model 106. The classification module 210 is configured to classify the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score.
The server 104 may optionally include any other modules and/or components known in the art to enable the institution in the comprehensive evaluation of the ESGperformance of the client. In an embodiment, the server 104 may include an Environmental, Social, and Governance - Wisdom, Intelligence, and Support Engine designed to facilitate interactions between multiple artificial intelligence (AI) agents, enabling them to collaborate and communicate effectively.
FIG. 3 is an architectural diagram that illustrates a multi-agent workflow of the system 100 of FIG.1 according to some example embodiments herein. These AI agents are powered by foundation models 302such as Chat-GPT 3/4, Llama-2, and like advanced AI models.The system 100 creates a unified platform where these agents can seamlessly work together, enhancing their collective capabilities. The knowledge base 306 serves as a repository for extensive question-and-answer functionality. Users can seamlessly add and scale a vast number of documents, including research reports, regulatory filings, sustainability disclosures, industry standards, and best practices. This scalability enables the system 100 to cover a wide range of ESG analysis topics and domains. The comprehensive repository enhances the platform's ability to deliver accurate, relevant, and current information in response to user queries. Continuous updates and curation ensure the inclusion of the latest data, research, and industry developments, maintaining the relevance and reliability of the system’s information offerings. The multi-agent workflow 304 facilitates autonomous agent collaboration to achieve common goals. This architecture features a persona-driven workflow, where specialized agents within the central application perform tasks related to ESG analysis.
FIG. 4 is an exemplary structural diagram for a multi-prompt engine of the system 100 of FIG.1 according to some example embodiments herein.The multi-agent engine 404 is integrated with vectorized data 402and data parser 406. The vectorized data is transformed into vector form for easier manipulation and analysis in machine learning and AI models, while a data parser extracts and formats raw data into a structured form suitable for processing. The system 100 includes an Environmental, Social, and Governance (ESG) specific knowledge base, which is a repository of information related to ESG topics, and includes but is not limited to, data, trends, regulations, best practices, and case studies. Integration with the ESG-specific knowledge base facilitates providing insights and information related to sustainability and responsible business practices. The multi-agent engine 404 is capable of plugging in and utilizing various language models, including, but not limited to, Chat-GPT 3/4 408, Google Bard 410, and LLama2 412, and potentially even more advanced models like future iterations of GPT, thereby allowing flexibility and adaptability. Different language models bring unique strengths and nuances to the AI agents’ interactions. The multi-prompt engine's architecture involves several key components as below:
Agent Management: The engine oversees the operation of multiple AI agents, managing their tasks, interactions, and coordination.
Communication: The AI agents can exchange information and insights through a structured communication protocol. This ensures that the agents collaborate effectively and share their expertise when required.
Prompt Formulation: The engine generates prompts that guide the AI agents' interactions. These prompts are carefully crafted to solicit specific responses that align with the objectives of the interaction.
Knowledge Integration: The ESG-specific knowledge base is integrated into the engine, allowing AI agents to access accurate and up-to-date information related to sustainability and responsible business practices.
Response Aggregation: The responses generated by individual AI agents are aggregated and synthesized to provide comprehensive and well-rounded answers to user queries or tasks.
Learning and Adaptation: The engine may incorporate machine learning techniques to improve over time. It can learn from successful interactions and user feedback to enhance its performance in delivering relevant and accurate information.
Thus, the system 100 provides the users with a holistic perspective on ESG matters, combining insights from different AI agents with diverse expertise. The collaborative approach helps in delivering comprehensive and well-informed responses. Moreover, the flexibility to incorporate various language models ensures that the engine remains adaptable to changing AI technologies and their capabilities.
The ESG-specific knowledge base of the system 100 and the capability to integrate various language models represent a cutting-edge solution for enabling advanced AI agents to collaborate effectively. By harnessing the power of multiple AI models and specialized knowledge, the engine provides valuable insights and information in the realm of sustainability and responsible business practices.
FIG. 5 is an exemplary structural diagram for an enablement-specific multi-agent framework of the system of FIG. 1 according to some example embodiments herein. The enablement-specific multi-agent framework illustrates a sophisticated system centered around a Multi-Agent LLM Engine 502. This engine encompasses a Recommendations/Insight Agent 504, which is further segmented into three specialized agents: The Data Analytics Agent 506, Summary Agent 508, and Citation Agent 510. The Data Analytics Agent 506 includes a Chart Builder 512, Savings Calc 514, and Carbon Calc 516 to handle data visualization and calculations. The Summary Agent 508 features a Knowledge Base 518, Research Papers 520, and Web Search 522 to aggregate and synthesize information. The Citation Agent 510 manages Client Data 524, BLBI Data 526, and Public Data 530 to ensure comprehensive and accurate citations. This framework is designed to leverage the combined capabilities of these agents to deliver actionable insights and recommendations.
The multi-agent architecture represents a cutting-edge solution for enabling advanced AI collaboration in the realm of sustainability, empowering users with the tools and knowledge to drive positive environmental, social, and governance outcomes.
Data analytics enables organizations to identify ESG trends, assess performance against sustainability goals, and quantify the financial implications of ESG initiatives. By analyzing ESG data, companies can make informed decisions to improve their environmental footprint, enhance social impact, and strengthen governance practices.
ESG scenario testing helps organizations anticipate and prepare for ESG-related challenges, such as climate change, supply chain disruptions, and regulatory shifts. By stress-testing their ESG strategies against various scenarios, companies can enhance resilience, mitigate risks, and seize opportunities for sustainable growth.
ESG research provides insights into companies' sustainability practices, performance, and impact. Investors, analysts, and other stakeholders use ESG research to assess risks, inform investment decisions, engage with companies on ESG issues, and drive positive change toward more sustainable business practices.
ESG report generation enables companies to transparently communicate their ESG performance, initiatives and impacts to stakeholders. Well-designed ESG reports enhance accountability, credibility, and trust, driving investor confidence, regulatory compliance, and stakeholder engagement in sustainability efforts.
FIG. 6 illustrates a sample interface of a company’s dashboard, detailing its ESG profileaccording to some example embodiments herein. As a part of the company profile, the dashboard includes theCompany ESG Profile, Public ESG Sentiment & Controversies, ESG KPI breakdown, ESG performance Insights, and Peer Benchmarking. The system features a repository of the top 1000 Listed Indian companies' ESG profiles and also can generate an ESG assessment of a new company on request.In addition, it provides a general ESG chat section forinquiries related to sustainability.
Based on an analysis of existing public ESG reports and relevant publications, the system provides comprehensive peer and industry benchmarking for various companies. This approach includes detailed ESG KPI comparisons across industry peers, evaluating both ESG performance and controversy scores. This allows to provide a holistic assessment that aids companies in understanding their relative standing within their industries and among their peers.
FIG. 7 is a flow diagram that illustrates a methodfor generating a machine learning generated score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score according to some embodiments herein. At step 702, the method includes obtaining ESG-related data including annual reports, ESG reports, and sustainability reports of the client from a plurality of data sources and peer clients and a public sentiment data related to the clientfrom media and public reports. At step 704, the method includes generating using the AI model, (i) a first machine learning score by analyzing the ESG-related data of the client, and (ii) a second machine learning score by analyzing the public sentiment data of the client. At step 706, the method includes generating an aggregate machine learning-generated score based on the first machine learning-generated score and the second machine learning-generated score. At step 708, the method includes classifying the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score. The aggregate machine learning generated score is equal to the first machine learning generated score when the second machine learning generated score is greater than or equal to the first machine learning generated score and the aggregate machine learning generated score is an average of the first machine learning generated score and the second machine learning generated score when the second machine learning generated score is lesser than the first machine learning generated score.
The first machine learning-generated score is determined by (i) extracting and analyzing data points from the ESG-related data, (ii) assigning KPIs to categories comprising environment, social and governance based on themes, (iii) assigning the categories to ESG pillars, wherein the ESG pillars comprise environmental pillar, social pillar, and governance pillar, (iv) normalizing the data points by standardizing KPI values, wherein the KPI values are standardized by dividing each KPI value by the client’s revenue for the fiscal year, (v) determining the percentile score for each data point, (vi) summing individual data point scores to obtain category scores, (vii) determining a weighted average for each category based on weights, wherein the weight for the environment pillar and the social pillar comprises 45% and the weight for the governance pillar comprises 10%, (viii) determining the first machine learning score by summing the weighted average for each category.
The second machine learning score is determined by (i) performing sentiment analysis on the public sentiment data using large language models (LLMs) to classify the public sentiment data into positive, negative, and neutral categories, (ii) assigning a score of 100 to clients with no identified controversies, (iii) applying severity weights to controversy counts for addressing market cap bias, wherein the severity weights comprises 0.33 for large-cap clients, 0.67 for mid-cap clients and 1 for small-cap clients, (iv) determining the count of controversies per client and multiplying the count of controversies by the severity weight based on the client’s market cap class, and (v) applying a pre-defined percentile scoring methodology to derive the second machine learning score. The predefined percentile scoring methodology implements a plurality of factors comprising a number of peer clients with values than the client, a number of peer clients with the same value as the client, and a number of peer clients with no values for a corresponding KPI or ESG-related metric.
The method includes generating (i) a first survey questionnaire for the client, and (ii) a second survey questionnaire for the client, wherein the first survey questionnaire comprises a set of questions based on ESG categories, wherein the second survey questionnaire comprises a set of questions based on ESG Controversies.
The ESG categories include Business Model Resilience, Materials Sourcing & Efficiency, Physical Impacts of Climate Change, Product Design & Lifecycle Management, Supply Chain Management, Air Quality, Ecological Impacts, Energy Management, GHG Emissions, Waste & Hazardous Materials Management, Water & Wastewater Management, Employee Engagement, Diversity & Inclusion, Employee Health & Safety, Labor Practices, Business Ethics, Competitive Behavior, Critical Incident Risk Management, Management of the Legal & Regulatory Environment, Systematic Risk Management, Access and Affordability, Human Rights & Community Relations, Product Quality and Safety, Selling Practices and Customer Welfare, Data Security, and Human Rights & Community Relations.
The ESG controversies include Anti-competition controversies, Business ethics controversies, Intellectual property controversies, Critical countries controversies, Public health controversies, Tax fraud controversies, Child labour controversies, Human rights controversies, Management compensation controversies, Consumer controversies, Customer health and safety controversies, Privacy controversies, Product access controversies, Responsible marketing controversies, Responsible R&D controversies, Environmental controversies, Accounting controversies, Insider dealings controversies, Shareholder rights controversies, Diversity and opportunity controversies, Employee health and safety controversies, Wages or working condition controversies, and Strikes.
In some embodiments, the method includes (i) presenting the first survey questionnaire and the second survey questionnaire to the first computing device associated with the client, and (ii) determining a first self-assessment score of the client based on the responses received from the client to the first survey questionnaire, and a second self-assessment score based on the responses received from the client to the second survey questionnaire.
The method includes determining a deviation between the first self-assessment score, the second self-assessment score, and the aggregate machine learning score, wherein the deviation indicates a disparity between the client’s perception and the assessed ESG risk levels.
The ESG self-assessment questionnaire is based on a Likert scale wherein each choice represents a different level of performance or compliance. A 5-point Likert scale survey may be generated, wherein the participants are provided with 5 answer choices, each representing a different level of performance or compliance. Scores ranging from 1 to 5 are assigned to each answer thus allowing for a comprehensive subjective evaluation of the client's ESG performance.
An example embodiment of the ESG self-assessment questionnaire based on the 5-point Likert scale is shown herein below:
Question: Please select the statement that best illustrates the level of ESG responsibility within your organization:
Board members have I&D representation and are responsible for actively driving strategic ESG initiatives; CEO and C-Suite have one or more ESG objectives tied to financial compensation
Board member(s) responsible for strategic ESG initiatives but play a passive role; CEO and C-Suite responsible for ESG but have no ESG objectives tied to financial compensation
C-suite or Management responsible for ESG with limited visibility of topic on board-level
None of the Above
Don't Know
The Likert-scoring system enables a nuanced assessment of the client's sustainability practices, facilitating targeted improvements and highlighting areas of strength and potential growth.
An example embodiment of the questionnaire of the ESGcontroversy assessment survey is shown herein below:
Question: On a scale of 1 to 5, how would you rate the frequency of strikes or industrial disputes that have led to lost working days in our company?
Never
Rarely
Sometimes
Often
Almost always
Following the findings from the ESG controversy assessment survey, the ESG controversy score is determined. This score is based primarily on the number of controversies attributed to each client within a specified financial year. No controversy is double-counted. The default value for all controversy measures is set to 0. Theclients with no controversy areassigned a score of 100 indicating their strong performance. Thereafter, the systemmultiplies the count of controversies by severity weights for eachclient, based on its market cap class. Severity weights are applied to address market cap bias, which often affects large companies, and areapplicable for the calculation of current and historical periods. Larger clients tend to attract more media attention, so to ensure a fair evaluation of the ESG controversy score across all client sizes, the severity weights are applied as per the market cap class as shown in the followingTable No. 1.
Global benchmark Cap Class Severity rate
>=10 billion Large 0.33
>=2 billion Mid 0.67
<2 billion Small 1
Table No. 1
Thereafter, the values obtained after applying the severity weights, sorts the clientsin ascending order with the lowest value representing better performance.
The percentile rank scoring methodology is utilized to determine the ESGcontroversy score by considering, (a) how many companies are lower or worse ESG score than the current one, (b) how many companies have the same value, and (c) how many companies have a value at all, that is, it determines the total number of companies included in the comparison set that have any value for the ESG controversy score, regardless of whether they are better, worse, or equal to the current company.The system determinesthe ESGcontroversy score by applying the percentile rank scoring formula as follows:
ESG controversy score = No. of companies with worst value + No. of companies with the same value/2
No. of companies with value
The systemthereafter compares the ESG score obtained by subjective evaluation from the ESG self-assessment surveyand the final ESG score (by objective evaluationbased on standardized criteria, and benchmarked against peers). Any deviation between the ESG score from the ESG self-assessment survey and the final ESG score provides valuable insights into the client'sperceptions of its performance versus its actual ESG performance. Analyzing the reasons for such deviation may help identify the client’s strengths, weaknesses, and areas for improvement thereof.
For example, the combined score logics is shown in the following Table No. 2.
Scenario
ESG controversies
score ESG score Final ESG score
1 100 89 89
2 48 49 48.5
Table No. 2
A higher ESG score of the client than that of the peer group of companies indicates strong ESG performance of the client and highlights areas of strengths that may offer a competitive advantage and serve as a valuable differentiator to help attract investors, customers, and other stakeholders prioritizing sustainability and responsible business practices. Whilst, a lower ESG and ESG controversy score of the client as compared to the peer group of companies indicates low ESG performance of the client and highlights areas of potential weakness or risks where improvements are required. Highlighting these areas of underperformance facilitates the client in strategizing and prioritizing their sustainability practices and allocating resources more effectively to address the identified gaps, risks, and challenges, thereby enabling the client to mitigate potential negative impacts and enhance their ESG performance.
For determiningthe ESG score, the reports such as annual reports, ESG reports, and sustainability reports published by the company are evaluated. The logic uses the 92 identified KPI’s along with their polarity and sections. From these base kpi we build up to categories and finally to pillars. The following Table No. 3lists the KPIs used in the method.
KPI POLARITY SECTION CATEGORY PILLAR
Total electricity consumption in Joules from renewable sources P Energy Resource Environment
Total fuel consumption in Joules from renewable sources P Energy Resource Environment
Total Energy consumption in Joules from renewable sources P Energy Resource Environment
Total electricity consumption in Joules from non-renewable sources N Energy Resource Environment
Total fuel consumption in Joules from non-renewable sources N Energy Resource Environment
Total Energy consumption in Joules from non-renewable sources N Energy Resource Environment
Water withdrawal by source (in kiloliters) from surface water P Water Resource Environment
Water withdrawal by source (in kiloliters) from groundwater P Water Resource Environment
Water withdrawal by source (in kiloliters) from third party water N Water Resource Environment
Amount of rainwater was utilized (in kiloliters) P Water Resource Environment
Volume of water withdrawal (in kiloliters) N Water Resource Environment
Total volume of water consumption (in kiloliters) N Water Resource Environment
Water intensity per rupee of turnover (Water consumed / turnover) N Water Resource Environment
Total Scope 1 emission in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 2 emission in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 1 and Scope 2 emissions per rupee of Turnover N Analytic CO2 Emission Environment
Total Scope 3 emission in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the category 1 – Purchased goods & services in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the Category 2 – Capital goods in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the Category 3 Fuel and energy related activities (not included in Scope
1 or 2)s in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the category 4 – Upstream transportation and distribution in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the Category 5 – Waste generated in operations in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the Category 6 – Business travel in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emission under the Category 7 – Employee commuting in metric tonnes of CO2 N Analytic CO2 Emission Environment
Total Scope 3 emissions per rupee of
Turnover N Analytic CO2 Emission Environment
Proportion of female employees among the total workforce P Structure (independence, diversity,
committees) Management Governance
Percentage of differently-abled male employees P Diversity and Inclusion Workforce Social
Percentage of differently-abled female employees P Diversity and Inclusion Workforce Social
Percentage of female representation on the board of directors P Structure (independence, diversity, committees) Management Governance
Percentage of female representation in the key management personnel P Structure (independence, diversity, committees) Management Governance
Percentage of female representation in the Senior management P Structure (independence, diversity, committees) Management Governance
Total turnover rate for permanent employees N Compensation Management Governance
Total number of complaints filed from communities N Human Rights Human Rights Social
Total number of
Complaints
Filed from investors other than shareholders N Human Rights Human Rights Social
Total number of complaints filed from shareholders N Human Rights Human Rights Social
Total number of complaints filed from customers N Human Rights Human Rights Social
Total number of complaints filed from employees N Human Rights Human Rights Social
Total number of complaints filed from value chain partners N Human Rights Human Rights Social
Percentage of R&D investments in specific technologies P Green revenues, research and
development (R&D) and capital
expenditures (CapEx) Innovation Environment
Percentage of capital expenditure (capex) investments in specific technologies P Green revenues, research and
development (R&D) and capital
expenditures (CapEx) Innovation Environment
Percentage of total employees covered by health insurance P Compensation Management Governance
Percentage of total employees covered by accident insurance P Compensation Management Governance
Percentage of total employees covered by maternity benefits P Compensation Management Governance
Percentage of total employees covered by paternity benefits P Compensation Management Governance
Percentage of total employees covered by day care facilities P Compensation Management Governance
Percentage of Return to work of permanent male employees that took parental leave P Compensation Management Governance
Percentage of Return to work of permanent female employees that took parental leave P Compensation Management Governance
Percentage of Return to work of total permanent employees that took parental leave P Compensation Management Governance
Percentage of Retention rates of permanent male employees that took parental leave P Compensation Management Governance
Percentage of Retention rates of permanent female employees that took parental leave P Compensation Management Governance
Percentage of Retention rates of total permanent employees P Compensation Management Governance
Percentage of total employees were given training on health and safety measures P Career development and training Workforce Social
Percentage of total employees were given training on skill upgradation P Career development and training Workforce Social
Lost time injury frequency rate (ltifr) N Health and Safety Workforce Social
Total recordable work-related injuries N Health and Safety Workforce Social
Total No. Of fatalities N Health and Safety Workforce Social
Number of High consequence work-related injury N Health and Safety Workforce Social
Number of employees having suffered high consequence work-related injury / ill-health / fatalities N Health and Safety Workforce Social
Number of employees who have been rehabilitated and placed in suitable employment or whose family
Members have been placed in suitable employment P Health and Safety Workforce Social
Percentage of total employees who have been provided training on human rights issues and policy(ies) P Human Rights Human Rights Social
Percentage of male permanent employees who are paid with minimum wage N Human Rights Human Rights Social
Percentage of female permanent employees who are paid with minimum wage N Human Rights Human Rights Social
Percentage of male permanent employees who are paid with more than minimum wage P Human Rights Human Rights Social
Percentage of male permanent employees who are paid with more than minimum wage P Human Rights Human Rights Social
Number of Complaints on Sexual harassment by employees N Human Rights Human Rights Social
Number of Complaints on Child Labour N Human Rights Human Rights Social
Number of Complaints on Discrimination at the workplace by employees N Human Rights Human Rights Social
Number of Complaints on Forced Labour/Involuntary Labour by employees N Human Rights Human Rights Social
Number of Complaints on Wages N Human Rights Human Rights Social
Number of Complaints on Other human rights related issues by employees N Human Rights Human Rights Social
Number of consumer complaints in data privacy N Data privacy Product
Responsibility Social
Number of consumer complaints in Advertising N Responsible Marketing Product
Responsibility Social
Number of consumer complaints in Cyber-security N Data privacy Product
Responsibility Social
Number of consumer complaints in Delivery of essential services N Responsible Marketing Product
Responsibility Social
Number of consumer complaints in Restrictive Trade Practices N Responsible Marketing Product
Responsibility Social
Number of consumer complaints in Unfair Trade Practices N Responsible Marketing Product
Responsibility Social
Total amount of waste generated in metric tonnes N Waste Emission Environment
Total amount of plastic waste generated N Waste Emission Environment
Total amount of E-waste generated N Waste Emission Environment
Total amount of Biomedical waste generated N Waste Emission Environment
Total amount of Construction and demolition waste generated N Waste Emission Environment
Total amount of Battery waste generated N Waste Emission Environment
Total amount of Radioactive waste generated N Waste Emission Environment
Total amount of Other hazardous waste generated N Waste Emission Environment
Total amount of Other non-hazardous waste generated N Waste Emission Environment
Total amount of waste recovered through recycling, re using or other recovery operations P Waste Emission Environment
Total amount of total waste disposed P Waste Emission Environment
Total Energy consumption in litres from diesel N Energy Resource Environment
Total Energy consumption in litres from petrol N Energy Resource Environment
Total number of employees in the age group of 18-30 P Diversity and Inclusion Workforce Social
Total number of employees in the age group of 31-50 P Diversity and Inclusion Workforce Social
Economic value distributed P Compensation Management Governance
Table No. 3
Standardizing KPIs involves dividing their values by the company's revenue (in billion USD). This normalization process allows for a fair comparison and assessment of each metric's impact relative to the company's size.
Normalized KPI value=(KPI value)/(Revenue of the company for that fiscal year)

The system allows users to add or remove KPIs based on the relevance and importance to their assessment objectives. For example, if a user wants to focus more on environmental factors, they can add additional KPIs related to carbon emissions, renewable energy usage, or waste management. Similarly, if certain KPIs are deemed less relevant or redundant, they can be removed to streamline the assessment process.
The individual data points score is summed up to get the category score. The weighted average is shown in Table No. 4 below:
Category Subcategory Weight (%)
Environment(45) Water 22.5
Waste 22.5
Energy 22.5
Analytic CO2 22.5
Green Revenue 10
Total: 100
Social(45) Diversity and Inclusion 20
Human Rights 20
Career Development 20
Health and Safety 10
Data Privacy 20
Responsible Marketing 10
Total: 100
Governance(10) Structure 50
Compensation 50
Total: 100
Table No. 4
ESG score is determined as follows:
ESG = (0.45*Environmental score)+(0.45*Social score)+(0.45*Governance Score)
FIG 8. is a schematic diagram of a computer architecture according to some embodiments herein, with reference to FIGS. 1 through 8. This schematic drawing illustrates a hardware configuration of a system for generating machine learning scoreor computer system or the computing devices in accordance with the embodiments herein. The system includes at least one processing device CPU 10 that may be interconnected via system bus 14 to various devices such as random-access memory (RAM) 12, read-only memory (ROM) 16, and an input/output (I/O) adapter 18. The I/O adapter 18 can connect to peripheral devices, such as disk unit 38 and program storage devices 40 that are readable by the system. The system can read the inventive instructions on the program storage devices 40 and follow these instructions to execute the methodology of the embodiments herein.
The system further includes a user interface adapter 22 that connects a keyboard 28, mouse 30, speaker 32, microphone 34, and/or other user interface devices such as a touch screen device (not shown) to the bus 14 to gather user input. Additionally, a communication adapter 20 connects the bus 14 to a data processing network 42, and a display adapter 24 connects the bus 14 to a display device 26, which provides a graphical user interface (GUI) 36 of the output data in accordance with the embodiments herein, or which may be embodied as an output device such as a monitor, printer, or transmitter, for example.
Thesystemand the method facilitate the institutions to contribute to a sustainable future by providing a comprehensive and integrated approach toESG risk score generation and portfolio management. The system and the method enable the institutions involved in financial decision-makingto understand and manage the ESG riskscores associated with their lending portfolios. The system and the method may automatethe data acquisition and sorting process therebysaving time and effort for the institution andallowing them to quickly visualize and compare the ESG performance of the client and the peer group of companies to identify high-risk clients and take more informed lending decisions that have a positive impact on the planet. The system and the method compare the client’s ESG performance with the peer group of companies thereby enabling the institution to identify outliners and better access the client’s ESG riskscores. Comparative assessment with peer companies provides a benchmark for the client’s ESG performance and enables the institutions to make more informed lending decisions that align with their investment strategy.The system and the method use public sentiment analysis to highlight key sustainability controversies associated with the client and peer group of companies thus enabling the institution to better understand the client's reputation and potential risks associated with their lending decisions.
The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications without departing from the generic concept, and, therefore, such adaptations and modifications should be comprehended within the meaning and range of equivalents of the disclosed embodiments.It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments,those skilled in the art will recognize that the embodiments herein can be practiced with modification within the scope of the appended claims.
,CLAIMS:I/We claim:
1. A system (100) for generating a machine learning generated score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score, wherein the system (100) comprises:
a server (104), wherein the server (104) comprising,
a memory that includes a set of instructions; and
a processor that executes the set of instructions and is configured to
obtain (i) ESG-related data of the client from a plurality of data sources (102A, 102B, and 102N) and peer clients, wherein the ESG-related data comprises annual reports, ESG reports, and sustainability reports of the client and (ii) a public sentiment data related to the client from media and public reports;
generate using the AI model, (i) a first machine learning score by analyzing the ESG-related data of the client, and (ii) a second machine learning score by analyzing the public sentiment data of the client,
wherein the first machine learning generated score is determined by (i) extracting and analyzing data points from the ESG-related data, (ii) assigning Key Performance Indicators (KPIs) to categories comprising environment, social and governance based on themes, (iii) assigning the categories to ESG pillars, wherein the ESG pillars comprise environmental pillar, social pillar, and governance pillar, (iv) normalizing the data points by standardizing KPI values, wherein the KPI values are standardized by dividing each KPI value by the client’s revenue for the fiscal year, (v) determining the percentile score for each data point, (vi) aggregating individual data point scores to obtain category scores, (vii) determining a weighted average for each category based on weights, wherein the weight for the environment pillar and the social pillar comprises 45% and the weight for the governance pillar comprises 10%, and (viii) determining the first machine learning generated score by summing the weighted average for each category,
wherein the second machine learning-generated score is determined by (i) performing sentiment analysis on the public sentiment data using large language models (LLMs) to classify the public sentiment data into positive, negative, and neutral categories, (ii) assigning a score of 100 to clients with no identified controversies, (iii) applying severity weights to controversy counts for addressing market cap bias,(iv) determining the count of controversies per client and multiplying the count of controversies by the severity weight based on the client’s market cap class, and (v) applying a pre-defined percentile scoring methodology to derive the second machine learning generated score, and
generate an aggregate machine learning score based on the first machine learning generated score and the second machine learning generated score; and
classifying the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score, wherein the aggregate machine learning generated score is equal to the first machine learning generated score when the second machine learning generated score is greater than or equal to the first machine learning generated score, wherein the aggregate machine learning generated score is an average of the first machine learning generated score and the second machine learning generated score when the second machine learning generated score is lesser than the first machine learning generated score.

2. The system (100)as claimed in claim 1, wherein the server (104) is communicatively connected with afirst computing device associated with the client being a potential borrower and a second computing device (112) associated with an organization involved in financial decision-making, wherein the server (104) is configured to communicate the machine learning generated score to the second computing device (112) associated with the organization.

3. The system (100) as claimed in claim 1, wherein the KPIs are categorized and assigned a polarity based on their context, the polarity being either positive or negative.

4. The system (100) as claimed in claim 1, wherein the processor is configured to analyze negative public sentiment data using the large language models (LLMs) to identify specified controversial topics for determining the second machine learning generated score.
5. The system (100) as claimed in claim 1, wherein the severity weights comprises 0.33 for large-cap clients, 0.67 for mid-cap clients, and 1 for small-cap clients.

6. The system (100) as claimed in claim 1, wherein the predefined percentile scoring methodology implements a plurality of factors comprising a number of peer clients with lower values than the client, a number of peer clients with same value as the client and a number of peer clients with values for a corresponding KPI or ESG-related metric.

7. The system (100) as claimed in claim 1, wherein the processor is configured to generate (i) a first survey questionnaire for the client, wherein the first survey questionnaire comprises a set of questions based on ESG categories, and (ii) a second survey questionnaire for the client, wherein the second survey questionnaire comprises a set of questions based on ESG Controversies.

8. The system (100) as claimed in claim 7, wherein the ESG categories comprise one or more of Business Model Resilience, Materials Sourcing & Efficiency, Physical Impacts of Climate Change, Product Design & Lifecycle Management, Supply Chain Management, Air Quality, Ecological Impacts, Energy Management, GHG Emissions, Waste & Hazardous Materials Management, Water & Wastewater Management, Employee Engagement, Diversity & Inclusion, Employee Health & Safety, Labor Practices, Business Ethics, Competitive Behavior, Critical Incident Risk Management, Management of the Legal & Regulatory Environment, Systematic Risk Management, Access and Affordability, Human Rights & Community Relations, Product Quality and Safety, Selling Practices and Customer Welfare, Data Security, or Human Rights & Community Relations.

9. The system (100) as claimed in claim 7, wherein the ESG controversies comprise one or more of Anti-competition controversies, Business ethics controversies, Intellectual property controversies, Critical countries controversies, Public health controversies, Tax fraud controversies, Child labour controversies, Human rights controversies, Management compensation controversies, Consumer controversies, Customer health and safety controversies, Privacy controversies, Product access controversies, Responsible marketing controversies, Responsible R&D controversies, Environmental controversies, Accounting controversies, Insider dealings controversies, Shareholder rights controversies, Diversity and opportunity controversies, Employee health and safety controversies, Wages or working condition controversies, or Strikes.
10. The system (100) as claimed in claim 7, wherein the processor is configured to (i) present the first survey questionnaire and the second survey questionnaire on the first computing device associated with the client, and (ii) determine a first self-assessment score of the client based on the responses received from the client to the first survey questionnaire, and a second self-assessment score based on the responses received from the client to the second survey questionnaire.

11. The system (100) as claimed in claim 10, wherein the processor is configured to determine a deviation between the first self-assessment score, the second self-assessment score, and the aggregate machine learning generated score, wherein a deviation indicates a disparity between the client’s perception and the assessed ESG risk levels.

12. A method for generating a machine learning-generated score using an artificial intelligence (AI) model that indicates an ESG (Environmental, Social, and Governance) risk level of a client for classifying the client based on the machine learning generated score, wherein the method comprises:
obtaining (i) ESG-related data of the client from a plurality of data sources (102A, 102B, and 102N) and peer clients, wherein the ESG-related data comprises annual reports, ESG reports, and sustainability reports of the client and (ii) a public sentiment data related to the client from media and public reports;
generating using the AI model, (i) a first machine learning score by analyzing the ESG-related data of the client, and (ii) a second machine learning score by analyzing the public sentiment data of the client,
wherein the first machine learning generated score is determined by (i) extracting and analyzing data points from the ESG-related data, (ii) assigning KPIs to categories comprising environment, social, and governance based on themes, (iii) assigning the categories to ESG pillars, wherein the ESG pillars comprise environmental pillar, social pillar, and governance pillar, (iv) normalizing the data points by standardizing KPI values, wherein the KPI values are standardized by dividing each KPI value by the client’s revenue for the fiscal year, (v) determiningthe percentile score for each data point, (vi) summing individual data point scores to obtain category scores, (vii) determining a weighted average for each category based on weights, wherein the weight for the environment pillar and the social pillar comprises 45% and the weight for the governance pillar comprises 10%, (viii) determining the first machine learning score by summing the weighted average for each category,
wherein the second machine learning score is determined by (i) performing sentiment analysis on the public sentiment data using large language models (LLMs) to classify the public sentiment data into positive, negative, and neutral categories, (ii) assigning a score of 100 to clients with no identified controversies, (iii) applying severity weights to controversy counts for addressing market cap bias, wherein the severity weights comprises 0.33 for large-cap clients, 0.67 for mid-cap clients and 1 for small-cap clients, (iv) determining the count of controversies per client and multiplying the count of controversies by the severity weight based on the client’s market cap class, and (v) applying a pre-defined percentile scoring methodology to derive the second machine learning score, and
generating an aggregate machine learning score based on the first machine learning generated score and the second machine learning generated score, wherein the aggregate machine learning generated score is equal to the first machine learning generated score when the second machine learning generated score is greater than or equal to the first machine learning generated score, wherein the aggregate machine learning generated score is an average of the first machine learning generated score and the second machine learning generated score when the second machine learning generated score is lesser than the first machine learning generated score; and
classifying the client into high, medium, or low-risk levels on ESG criteria based on the aggregate machine learning generated score.

13. The method as claimed in claim 12, wherein the KPIs are categorized and assigned a polarity based on their context, the polarity being either positive or negative.

14. The method as claimed in claim 12, wherein the method comprises analyzing negative public sentiment data using the large language models (LLMs) to identify specified controversial topics for determining the second machine learning generated score.

15. The method as claimed in claim 12, wherein the predefined percentile scoring methodology implements a plurality of factors comprising a number of peer clients with values than the client, a number of peer clients with same value as the client, and a number of peer clients with values for a corresponding KPI or ESG-related metric.

16. The method as claimed in claim 12, wherein the method comprises generating (i) a first survey questionnaire for the client, and (ii) a second survey questionnaire for the client, wherein the first survey questionnaire comprises a set of questions based on ESG categories, wherein the second survey questionnaire comprises a set of questions based on ESG Controversies.

17. The method as claimed in claim 16, wherein the ESG categories comprise Business Model Resilience, Materials Sourcing & Efficiency, Physical Impacts of Climate Change, Product Design & Lifecycle Management, Supply Chain Management, Air Quality, Ecological Impacts, Energy Management, GHG Emissions, Waste & Hazardous Materials Management, Water & Wastewater Management, Employee Engagement, Diversity & Inclusion, Employee Health & Safety, Labor Practices, Business Ethics, Competitive Behavior, Critical Incident Risk Management, Management of the Legal & Regulatory Environment, Systematic Risk Management, Access and Affordability, Human Rights & Community Relations, Product Quality and Safety, Selling Practices and Customer Welfare, Data Security, and Human Rights & Community Relations.

18. The method as claimed in claim 16, wherein the ESG controversies comprise Anti-competition controversies, Business ethics controversies, Intellectual property controversies, Critical countries controversies, Public health controversies, Tax fraud controversies, Child labour controversies, Human rights controversies, Management compensation controversies, Consumer controversies, Customer health and safety controversies, Privacy controversies, Product access controversies, Responsible marketing controversies, Responsible R&D controversies, Environmental controversies, Accounting controversies, Insider dealings controversies, Shareholder rights controversies, Diversity and opportunity controversies, Employee health and safety controversies, Wages or working condition controversies, and Strikes.

19. The method as claimed in claim 16, wherein the method comprises determining (i) a first self-assessment score of the client based on the responses received from the client to the first survey questionnaire, and (ii) a second self-assessment score based on the responses received from the client to the second questionnaire.

20. The method as claimed in claim 19, wherein the method comprises determining a deviation between the first self-assessment score, second self-assessment score, and the aggregate machine learning generated score, wherein the deviation indicates a disparity between the client’s perception and the assessed ESG risk levels.

Dated this 31st day of July, 2024

Signature of Agent:
(Arjun Karthik Bala)
IN/PA-1021