I need an AI agent that can take over the bulk of our email and live-chat support. The system must be able to: • Answer routine questions about products, shipping, returns and policies. • Resolve or escalate complaints by following decision trees that I will supply. • Pull real-time order data and give customers up-to-the-minute tracking information. Personalization is a priority. Each reply should greet the customer by name, recognise previous purchases and reference earlier tickets so that follow-ups feel seamless. I can provide structured order data and historic conversation logs through our existing API endpoints. Please design, train and deploy the agent, then connect it to both channels (Zendesk and a simple web chat widget) so the hand-off from human to bot and back is smooth. I’d like clear documentation on how to retrain or extend the model later, plus a short demo session to verify performance. Acceptance criteria 1. Average reply time under three seconds. 2. At least 90 % intent-matching accuracy on a test set of 200 real tickets. 3. Full context retention across multi-turn conversations. 4. Easy toggle that lets my team switch the bot on or off per channel. If you have experience with GPT-4, Dialogflow CX, Rasa, or similar NLP frameworks as well as RESTful API integrations, you’ll be able to hit the ground running.