Soham Ghosh

About Soham Ghosh

Senior Product Manager building production-grade Generative AI (LLM) agents for enterprise customer service and CRM (Microsoft Dynamics 365). Shipped agentic workflows that automate case creation, triage, and resolution - delivering 95% reduction in manual effort and 90%+ reduction in average handling time in enterprise pilots. What I build LLM-powered AI agents / agentic systems for customer service and contact center automation End-to-end workflows: intent detection → reasoning → tool execution → resolution Copilot / assistant experiences with human-in-the-loop (HITL) controls Evaluation systems (LLM evals) to measure accuracy, safety, and business impact Enterprise-ready AI platforms with governance, observability, and ROI tracking Impact and scope Launched autonomous case management and email intelligence capabilities in Dynamics 365 Built AI agents for insurance workflows (FNOL, claim triage, resolution) Improved resolution accuracy and reduced false positives using large-scale feedback loops Enabled enterprise adoption via governance layers, transparency (Agent Hub), and ROI measurement Drove customer-zero deployments and multi-million dollar projected cost savings Core expertise Generative AI, Large Language Models (LLMs), AI Agents, Agentic Workflows, Retrieval-Augmented Generation (RAG), Prompt Engineering, Tool Use / Function Calling, Workflow Orchestration, LLM Evaluation (AI Evals), Human-in-the-loop (HITL), Responsible AI Practices, AI Governance, AI ROI Measurement, Customer Service AI, Agent Frameworks, LangChain, LangGraph, Microsoft Copilot Studio, n8n, Power Automate AI Builder, AI-First UX, AI Copilot / Assistant Design, Enterprise SaaS, CRMI love connecting with new people, you can reach me at LinkedIn or at sohamg2018(at)email.iimcal.ac.in.

Experience

1. Senior Product Manager

   Microsoft

   May 2024 — Present · Bengaluru, IN

   Working towards the vision of a fully autonomous Customer Service (CRM) transformed by Generative AI and Copilot experiences. Charter spans case management, email (as a support channel), service-level agreements, automatic record creation, and overall customer NSAT- Reduced manual case creation and update effort for pilot customers by 95% by launching out-of-the-box Agentic AI capabilities in D365 to extract data from support emails and chat- Reduced average handling time by 90%+ for 5 pilot customers by AI-driven intent extraction from support emails to auto-generate responses, deflect spam, trigger autonomous workflows- Reduced claim processing time by 80%+ for 3 pilot insurers by defining and launching an Insurance AI Agent on - Copilot Studio automating FNOL (First Notice of Loss) intake, claim triage, resolution- Accelerated adoption of AI agents by defining Agent Hub experience in D365, giving business leaders transparency, governance, ROI measurement, insights into Case Management Agent performance- Defined Model Context Protocol (MCP) specs for core case management scenarios, creating the foundation for plug-and-play AI agent connectivity with Dynamics 365- Unlocked projected $XXM annual cost savings driving customer zero adoption with Microsoft Support, piloting autonomous case creation, resolution, and closure in Dynamics 365- Improved AI case resolution accuracy by 15% and reduced false positives in spam deflection by incorporating feedback from 200+ Microsoft Support reps and supervisors across Power Platform, Surface, and other business lines before public launch- Drove migration of 100% of customers from legacy service level agreements and automatic record creation rules to modern versions by delivering a guided migration experience, best-practice documentation, and in-product tooling, ensuring smooth adoption at enterprise scale

Education

• Indian Institute of Management, Calcutta

  Master of Business Administration (M.B.A.)

  2016 — 2018

• University of California, Berkeley, Haas School of Business

  Berkeley Executive Education, Product Management

  2022 — 2022

• National Institute of Technology Durgapur

  Bachelor of Technology (B.Tech.), Electrical Engineering

  2010 — 2014