GENIE CRM: An Agentic AI-Powered Customer Relationship Management System with Multi-Modal Intelligence and Automated Workflow Orchestration
GENIE CRM: An Agentic AI-Powered Customer Relationship Management System with Multi-Modal Intelligence and Automated Workflow Orchestration
Author
Dev Tejas Shah
Department of Computer Science and Engineering,
Parul University, Vadodara, Gujarat, India
Guide
Prof. Hinaben Bharatbhai Dudharejiya
hinaben.dudharejiya33929@paruluniversity.ac.in
Department of Computer Science and Engineering, Parul University, Vadodara, Gujarat, India
April 2026
ABSTRACT:
The rapid evolution of large language models and agentic AI frameworks has opened a new frontier for intelligent enterprise software. This paper presents GENIE CRM, a full-stack customer relationship management platform that embeds autonomous AI agents across sixteen functional modules to eliminate manual bottlenecks throughout the sales, support, and operations lifecycle. Built on a Python-Flask backend and a React-TypeScript single-page application, the system leverages Google Gemini 1.5 Flash as a unified multimodal AI backbone to perform structured lead scoring, visiting card optical character recognition, personalised multi-channel outreach generation, intelligent support ticket classification and round-robin routing, document and call recording summarisation, AI-driven bug criticality assessment, and real-time geospatial business intelligence. Experimental evaluation on representative CRM workflows demonstrates a 94.4 percent reduction in lead data entry time through multimodal OCR, a 22.2 percentage-point improvement in automated ticket routing accuracy over manual methods, and a 97.3 percent structural success rate for AI-generated JSON responses across all service endpoints. A persistent voice-enabled chatbot named Genie provides conversational access to the CRM database from every page of the application. The architecture combines Supabase-backed PostgreSQL for relational storage with lightweight JSON flat files for rapid-iteration state management, achieving sub-1.2-second dashboard aggregation across all sixteen modules. This paper details the system architecture, module-level design methodology, UML activity and use case diagrams derived from actual source code analysis, quantitative evaluation results, design trade-off discussion, and a roadmap for cloud-native multi- tenant deployment.
Keywords — Agentic AI; Customer Relationship Management; Google Gemini; Lead Scoring; Optical Character Recognition; Workflow Automation; Support Ticket Routing; Geolocation Intelligence; Flask; React; Supabase; Multimodal AI; Sales Automation; Natural Language Processing; Chatbot