Authors: Priya Nair, Vikram Chauhan, Anika Deshpande, Vasudev Sharma
Abstract: Real time customer relationship management intelligence continues to evolve as organizations rely on advanced analytics to drive sales planning, revenue optimization, and customer engagement decisions. This study addresses persistent challenges related to data fragmentation, inconsistent contextualization of CRM information, and the limited adaptability of conventional predictive models within Salesforce environments. The research introduces a generative AI and large language model driven data fabric architecture designed to unify distributed CRM assets, automate semantic enrichment, and enhance predictive sales forecasting accuracy. A mixed methodological approach was adopted, combining architectural modeling, data flow simulation, and empirical evaluation using historical opportunity data, customer interaction logs, and multichannel engagement records. Findings indicate that the proposed model improves context aware forecasting precision, reduces data preparation overhead, and increases interpretability for frontline sales teams by enabling narrative style insights generated through domain tuned language models. The framework demonstrates the potential to streamline CRM operations, enhance cross system interoperability, and support adaptive decision making by integrating knowledge graphs and LLM based reasoning into the Salesforce ecosystem. The study contributes an extensible reference architecture for enterprise CRM analytics and offers a pathway for organizations seeking to modernize sales intelligence processes. The results hold significance for both practitioners and researchers by proving that next generation AI enabled data fabrics can meaningfully strengthen forecasting reliability, reduce operational friction, and support scalable data governance strategies across complex CRM landscapes.
DOI: http://doi.org/10.5281/zenodo.17894776
