I am building a fully-configurable AI agent for customer care. The core idea is a single bot that can 1) answer support tickets arriving via email, 2) handle live-chat conversations in real time, and 3) pull or push data by connecting to CRMs or any other external systems through their APIs. The bot must accept plain-language instructions so that non-technical staff can tweak tone, escalation rules, and knowledge sources without code changes. Under the hood I imagine an LLM-based architecture (GPT, Claude, or similar) orchestrated through a framework such as LangChain, Rasa, Botpress, or a custom Python/Node micro-service—whichever you believe fits best. Robust intent detection, conversation memory, and fallback to human agents are mandatory. I value experience in AI bot development above all, so please centre your proposal on that. Show me how you would: • design the conversation flow and knowledge retrieval layer • integrate with at least one popular CRM (HubSpot, Salesforce, Zendesk, etc.) via REST or GraphQL • expose configuration hooks so business users can add instructions or connect new data sources without redeploying the code Deliverables I expect: 1. Running prototype (Docker or cloud-hosted) covering the three key channels 2. Clean, well-documented code repository 3. Short admin guide outlining how to add new integrations and adjust behaviour Send a detailed project proposal that outlines milestones, tech stack, and timelines, plus any similar bots you’ve shipped before. I’m ready to move quickly once I see a clear plan.