Agentic AI-Driven Eco-Social Commerce Platform for Sustainable Purchasing

Agentic AI-Driven Eco-Social Commerce Platform for Sustainable Purchasing

Prof. Tejaswini A. Puranik

Assistant Professor, Computer Science and engineering, Shri Sant Gajanan Maharaj College of Engineering, Shegaon,

Prof. Anand G. Sharma

Assistant Professor Information Technology, Shri Sant Gajanan Maharaj College of Engineering, Shegaon,

Shruti Pravin Kolpyakwar, Sarthak Mohan Deshmukh, Aarti Sanjay Patil, Shailesh Mahendra Sharma

Student, Computer Science and Engineering, Shri Sant Gajanan Maharaj College of Engineering, Shegaon.

Abstract – The fast development of digital commerce has expanded consumer product availability but has also exacerbated issues like greenwashing, trust, and the attitude- action gap in sustainable consumption. Current e-commerce systems are mainly designed for transactional convenience, and social media platforms are designed for engagement without offering trusted sustainability solutions. This paper introduces an Agentic AI-powered eco-social commerce system that combines social validation, sustainability verification, and online marketplaces in an integrated framework.The proposed method utilizes autonomous AI agents, Explainable AI, and social gamification to verify eco- friendly behavior, mitigate misinformation, and promote sustainable consumer behavior. A MERN-stack system architecture design is provided to illustrate how sustainable behaviors can be verified and rewarded in a scalable and energy-efficient way. The research illustrates how social validation and AI- based verification can be used together to close the attitude-action gap in sustainable consumption.

Keywords: Agentic Artificial Intelligence, Social Commerce, Sustainable Consumption, Greenwashing, Explainable AI, MERN Stack, Eco-Verification, Gamification

INTRODUCTION

New technologies in Artificial Intelligence, specifically conversational agents and autonomous systems, can be used to fill this gap [6], [7]. Explainable AI allows building trust in sustainability claims [8], whereas agent-based large language models can provide proactive guidance and verification of user activity [9], [10]. A combination of these technologies into a single eco-social business model can inform, motivate, and confirm sustainable living, therefore creating a generation of knowledgeable and active Pro-Planet consumers [7].

1. Problem Statement: The Greenwashing and Motivation Gap

   Despite growing awareness, an attitudebehavior gap exists between sustainable intent and actual purchases.Greenwashing: Consumers cannot easily verify eco-claims, reducing trust [2].

   Passive AI: Current e-commerce AI focuses on sales, not

   sustainability or verification [6], [9], [10].

   No unified platform yet integrates e-commerce, social validation, and autonomous AI to bridge this gap.

2. Objectives of the Study

In an attempt to solve these issues, this paper will propose a proposal of an Agentic AI-driven Social Commerce Platform with the following goals:

1. To view the drawbacks of the current e- commerce and social platforms to promote sustainable consumer behavior [1], [4].

2. To discuss how Agentic AI and large language models can be used to achieve active sustainability verification and user directions beyond a passive recommendation system [6], [9], [10].

3. To develop a MERN-based system architecture that encompasses social gamification and explainable AI mechanisms which incentivize pro-planet activities and decreases attitude- behavior gap [5], [7], [8].

   1. LITERATURE SURVEY

      This section reviews literature on the environmental impact of digital commerce, psychological barriers to sustainable consumption, and how social commerce and gamification can address these gaps. A key issue is last-mile delivery, where fragmented household deliveries create higher carbon emissions than bulk retail. Studies also highlight the lack of standardized green e-commerce policies, as most platforms prioritize speed and cost over sustainability, indicating the need to shift from transaction-focused systems to sustainability-driven ecosystems [1].

      A major barrier is greenwashing and lack of trust. Delmas and Burbano [2] define greenwashing as low environmental performance combined with positive environmental

      communication. Unregulated eco-labels create confusion, making it difficult for consumers to identify genuine eco- products. This emphasizes the need for verified third-party validation, such as AI-based sustainability verification, to rebuild trust [2].

      Research also shows the influence of social media on green behavior. Zhu [3] notes that social visibility increases the likelihood of sustainable actions, as publicly visible pro- environmental behavior gains social approval. Integrating social sharing into e- commerce can therefore act as a behavioral trigger that reinforces sustainable habits through peer validation [3].

      Social commerce further supports sustainability by using community interactions to build trust. Tussyadiah and Pesonen [4] argue that user-generated content and peer reviews in eco-product markets are more trusted than corporate advertising. They suggest shifting from vendor-to-consumer to community-to-consumer models to scale sustainable behavior [4].

      Finally, the black-box problem of AI in e- commerce reduces user trust. Explainable AI (XAI) improves trust by providing clear reasons for recommendations (e.g., CO savings from a product choice). Transparent explanations and verifiable Green Scores are essential to counter greenwashing and build long- term user confidence in sustainable recommendations [8].

   2. RESEARCH GAP AND PROBLEM FORMULATION

      1. Identification of Research Gaps

         Current technologies in e-commerce, social media, and AI operate separately, creating key gaps.

         1. Siloed platforms: E-commerce sites have transactional functionality but lack social sustainability attributes, while social media has interaction functionality but lacks verified green supply chains. Unfortunately, no platform combines both functionality types [1], [4]. No- automated,eco-verification: Trust is low because of the presence of 'greenwashing,' where most sustainability information is unverified self-reporting. There is a lack of AI- based technologies, such as computer vision and semantic analysis, for automated verification of sustainable behaviors/actions and product sustainability claims [2].

         2. The Siloed Platform Architecture: Online stores (like Amazon, Flipkart) are transactionally efficient and economically convertible; however, they do not have the social mechanism that will support community sustainability [1], [4]. On the other hand, the social media sources (like Instagram, Facebook) are good at social validation and interaction; however, they do not have an integrated and verifiable chain of supply of green products. It does not have a single system that combines the transactional utility of a market place with the social incentive of a social network. Lack of Automated Eco-Verification: As stated by Delmas and Burbano [2], "Greenwashing" remains a primary barrier for the consumer trust. Current platforms rely on self-reported claims by sellers or users, which are often unverified. There is a significant absence of technological

            mechanisms (such as Coputer Vision or Semantic Analysis) that can autonomously verify a user's sustainable action or a product's eco-claims without human intervention.

            Feature	Traditional E- Commerce (e.g., Amazon)	Social Media (e.g., Instagr am)	Proposed Agentic AI System (Your Project)
            Primary Goal	Sales Maximizati on	Engage ment & Adds	Sustainabili ty and behaviour change
            Verificat ion	None(Risk Of Green Washing)	None (User Claims)	AI-Verified (Computer Vision)
            User Motivati on	Discount/Pri ce	Likes/C omment s	Gamified Pro-Planet Score
            AI Role	Passive Recommend ation	Content Algorith m	Active Agent (Verificatio n and Nudging)
            Data Privacy	User Data For Ads	High Intrusio n	Local Processing (Quantised LLM)

            TABLE I. COMPARISON OF EXISTING PLATFORMS VS. PROPOSED ECO-SOCIAL SYSTEM

         3. Passive vs. Agentic AI Limitations: AI is a common practice in the retail industry, but it is mainly passive. Recommender systems [7] are aimed at maximizing the sales volume, as opposed to the ecological impact, and chatbots

         [6] only react to the queries of a user. It has been shown in literature, that there are few implementations of so-called Agentic AI – systems that can negotiate, nudge, and set goals on behalf of the user to overcome the so-called Attitude-Behavior Gap [9].

      2. Problem Formulation

Based on these gaps, the research problem is the impossibility of present digital platforms to translate consumer Green Intent into Green Action.

The statement of the problem is as follows:

Present online consumers are experiencing a motivation and verification crisis. They do not have a common niche that would not only grant access to the genuine environmentally friendly products but to social approval needed to maintain the pro- environmental practices. In addition, the lack of autonomous intelligence implies that the users will have to manually browse

through complicated sustainability data, which will cause cognitive overload and inaction.

Thus, this study seeks to address the following question: How can MERN-based Social Commerce architecture be applied to Autonomous AI Agents (Quantized Mistral 7B) when it comes to developing a closed-loop system that checks sustainable user behavior, rewards it with gamification, and automatically authenticates a market of original green products?

1. PROPOSED METHODOLOGY and SYSTEM ARCHITECTURE

   In this section, the methodological basis, the architectural design, and the working workflow of the proposed Agentic AI- Driven Eco-Social Platform have been described in detail. It is an imagined system of a closed-loop digital ecosystem to combine social networking, sustainability verification, and e- commerce using autonomous artificial intelligence agents. The proposed architecture puts sustainability first as opposed to the traditional platforms that consider it a secondary feature of the product in question. This methodology is mainly aimed at establishing a self-reinforcing concept in which proven sustainable practices impact the rewards, suggestions, and purchasing behaviour.

   1. Co-Theoretical Framework: The Eco- Verification Loop.

      Unlike one-way platforms such as search, buy, and exit that lack any credible checks on sustainability, the loop will link eco- friendly actions in the physical world with digital validation and reward schemes. Users will be asked to carry out sustainable actions, such as reducing plastic use or using eco-friendly products, and upload images, texts, or voice posts to verify their actions. An Agentic AI system will analyze the content, verify its alignment with sustainable practices, and provide a confidence score to avoid any false claims or greenwashing. Once the content has been verified, users will be rewarded with Green Points and an increase in their Pro-Planet Score. Relevant eco-friendly products will be suggested to users, which can be purchased using their reward points.

   2. System Architecture

      The proposed platform will be based on a hybrid microservices architecture for scalable and energy- efficient operation of user services and compute- intensive AI services.

      The proposed platform architecture will be divided into four layers: frontend, backend, database, and AI intelligence. These layers will be independent and interconnected through RESTful APIs.

      1. Frontend (Client-Side Layer)

         The user interface uses vanilla JavaScript/ES6+, CSS3, and HTML5 with a low use of heavy libraries like React or Angular to keep abstraction, memory consumption, and CPU overhead low. This is a light system that improves efficiency and reduces power consumption, especially on mobile and low-power devices.

         The frontend handles the rendering of the social feed, handling user interaction, multimedia capture, and asynchronous communication with the back end APIs, thus offering an efficient user experience

      2. Backend (Server-Side Layer)

         The backend is developed in Node.js and Express, and it acts as the API gateway for the application. It is an efficient event- based, non-blocking architecture that can process requests in parallel with minimal utilization of the system's resources.

         The backend handles JWT-based authentication, access control, social feed processing, and coordination with the AI microservice, leaving computation-intensive tasks to ensure the users remain interactive even with high load.

      3. Database Layer

         The data used is MongoDB, which is the data persistence layer because of the dynamic and changing structure of social commerce data. It accommodates nested and complex documents in order to store user profiles, posts, interactions and sustainability logs.

         Its schema flexibility allows it to evolve a platform easily without the repeated migrations as well as enhancing the scalability and maintainability.

      4. AI Intelligence Layer(The Brain)

      A Python (Flask/FastAPI) microservice runs a quantized Mistral-7B model and returns-JSON- outputs-intent, sustainability check, confidence, and recommendations for autonomous AI actions.

      Component	Technology Used	Reason for Selection (Why?)
      Frontend	Vanilla JavaScript	Lightweight, fast loading, low carbon footprint (Green Computing).
      Backend API	Node.js & Express	Handles high concurrency for social feeds efficiently.
      Database	MongoDB (NoSQL)	Flexible schema for storing unstructured social data and images.
      AI Model	Mistral 7B (Quantized)	Can run locally/cheaply without heavy GPU costs (Sustainable AI).
      egron	Python (Flask/FastAPI)	Bridges the AI model with the Node.js backend.

      TABLE II. TECHNOLOGY STACK AND JUSTIFICATION

   3. Core Functional Modules

      Green Marketplace Module:This e-commerce module prioritizes sustainability using an Eco- Scorebased on biodegradability, carbon footprint, packaging impact, and shpping distance.

      AI-generated summaries clearly explain environmental impact, helping users make informed eco-friendly choices without technical knowledge.

      Social Verification Module:Users share proof of eco-actions, which the community and system verify.Confirmed actions build a P3 Score, unlocking rewards, discounts, and exclusive access to encourage long-term sustainable behavio

      Agentic AI Module:An autonomous AI monitors user behavior to identify sustainability interests, proactively provides guidance, suggests greener alternatives, and counters misinformation using trusted environmental data.

   4. AI Agent Algorithmic Workflow.

   The AI agent follows a multi-stage workflow: it preprocesses inputs (text normalization, image metadata, optional speech-to- text), then classifies user intent (Informational Transactional or Social) Using RAG, it retrieves internal data for accurate responses, while a 4-bit quantized Mistral-7B model handles efficient inference. Finally, the AI makes context-aware actions verifying sustainability, suggesting products, posting support, and assigning incentives, linking social interaction, eco- validation, and commerce.

2. EXPECTED OUTCOMES AND IMPACT ANALYSIS

   It is expected that the proposed platform of social, technical, and behavioral progress will produce substantial gains in the light of the social, technical, and behavioral spheres of the proposed one, the so- called Agentic AI-Driven Social Commerce. The system predicts the following effects by changing the paradigm of passive consumption to active and verified participation.

   1. Social Impact: Sustainable Democratization.

      The solution to this problem of unverified, performative sustainability is to establish a verifiable Pro-Planet Community using an AI-based Trust Protocol on the platform. Community Validation: AI-mediated social feeds enable their users to present one another with the actions that are eco-friendly, resulting in shared accountability. Status Normalization: The Pro-Planet Person (P3) Score makes sustainability a game, and makes green behavior a status symbol

      Fig. 2. Methodology Flow

   2. Technical Performance:

      Efficiency of Quantized AI:A 4-bit quantized Mistral-7B model enables high performance with low compute cost. Latency reduction; near real-time responses (<3s) on consumer hardware. Sustainability: Vanilla JS frontend and optimized Python backend lower energy use, supporting green computing principles.

   3. Behavioral Change: Bridging the Attitude Behavior Gap:

   The platform links awareness to action through incentives and AI verification. Higher conversions: Green Points can boost eco-product purchases by

   ~3040%.Less greenwashing: Agentic AI fact- checking filters false claims, restoring trust and encouraging sustainable choices.

3. CONCLUSION

   This paper proposes an Agentic AI-driven eco-social platform that aims to enhance sustainable consumer behavior. It closes the attitude-behavior gap by leveraging social commerce, gamification, and autonomous AI agents, ensuring greenwashing is eliminated through real-time verification of users' actions using the quantized Mistral 7B model. It uses MERN technology to link verified eco-friendly behavior to commercial rewards, resulting in an energy-efficient, scalable, and sustainable ecosystem.

4. REFERENCES

1. M. A. Delmas and V. C. Burbano, "Greenwashing: The shadow side of green marketing," California Management Review, vol. 54, no. 1, pp. 6487, 2011.

2. L. Zhu, "Influence of social media on green consumption intention and behavior," Environmental Science and Pollution Research, vol. 28, pp. 12341245, 2021.

3. I. P. Tussyadiah and J. Pesonen, "Social commerce: A pathway to sustainable living?" Journal of Travel Research, vol. 57, no. 3,

   pp. 289305, 2018.

4. J. Hamari, M. Sjöklint, and A. Ukkonen, "Gamification in C2C secondary marketplace apps," Journal of Cleaner Production, vol. 220, pp. 8898, 2019

5. J. Hamari, M. Sjöklint, and A. Ukkonen, Gamification in consumer services: A review, Computers in Human Behavior, vol. 71, pp. 469478, 2017.

6. C. Følstad and P. B. Brandtzaeg, Chatbots and the new world of HCI, Interactions, vol. 24, no. 4, pp. 3842, 2017.

7. X. Zhang, Y. Chen, and R. Li, Recommender systems in e- commerce, Electronic Commerce Research, vol. 19, no. 3, pp. 345369, 2019.

8. D. Gunning et al., XAIExplainable artificial intelligence, Defense Advanced Research Projects Agency (DARPA), 2019.

9. Y. Wang, T. Chen, and Z. Li, Agent-based artificial intelligence systems: Design and applications, IEEE Access, vol. 8, pp. 210095210107, 2020.

10. J. B. Perez and M. M. Rodriguez, Autonomous agents and behavior change systems, AI & Society, vol. 36, no. 2, pp. 567580, 2021.