The ECOSH Agentic AI Sales Assistant project aims to develop a modular AI system integrated into the ECOSH CRM platform to improve analytics, automation, and decision support in B2B sales workflows. Traditional CRM systems often suffer from fragmented data, limited automation, and inefficient analytics, while monolithic Large Language Models (LLMs) introduce risks such as hallucination, high computational cost, and poor domain adaptation.
The project proposes a modular agentic AI architecture where a central orchestration layer coordinates specialized AI components for Text-to-SQL generation, predictive analytics, and AI-assisted content generation. Retrieval-Augmented Generation (RAG) is used to ground outputs on verified enterprise data and improve reliability in business-critical workflows.
The system is designed as a reusable enterprise AI layer that can support additional business applications beyond the ECOSH CRM ecosystem. The project focuses on multilingual Turkish–English workflows, distributed AI inference, and scalable enterprise automation using GPU-accelerated infrastructure.
Expected outcomes include improved AI-assisted CRM analytics, reduced hallucination risks, lower operational AI costs, and validation of a scalable enterprise AI workflow suitable for SMEs and industrial environments.
Principal Investigator, Company and Country
Alya Rahik, TDG Bilişim Danışmanlık ve Müşteri Hizmetleri Ticaret A.Ş., Türkiye