Replace the email chain for product sample requests: let sales staff ask in plain words, and have an AI agent check stock, demand forecasts and past orders before it approves or flags each request: AI agent prompt

Help me set up simple, rule-based approvals for product sample and small order requests that an AI assistant can check against my stock data. My business: [what we sell, sales channels]. Data I can export: [e.g. inventory on hand by item, open orders, last 3 months of sales, demand forecast]. Who requests samples: [sales reps / marketing / partners]. 1) Write a short plain-language request form a rep can fill in by typing one sentence (e.g. 'Send 6 units of item X to retailer Y by Friday for a buyer meeting'), and list what the AI must pull out of it: item, quantity, recipient, purpose, deadline, requester. 2) Write approval rules, for example: approve automatically if the quantity is under [number] units and stock after the request stays above [number] weeks of forecast demand; send to a manager if the item is low, on backorder, new or high-value, or if the same customer received samples in the last [number] days. 3) Using my export [paste], check these sample requests [paste] against the rules and return a table: request, decision (approve / needs manager / reject), the numbers behind it in one line, and a draft reply to the requester. 4) Suggest a monthly summary showing samples sent by rep, customer and item, and their cost. You only recommend decisions; a person releases every shipment, and you never change stock or orders.