AI Automation Engineer to Build an Intelligent Order & Inquiry Processing System

We're a mid-sized wholesale supplier based in Melbourne, distributing café and hospitality equipment to around 400 recurring B2B customers across Australia. Most of our orders and inquiries still come in by email, and a few by PDF purchase order attachments. Right now two of our staff spend most of their day reading inbound emails, working out whether each one is a new order, a reorder, a quote request, a delivery query, or a complaint, then manually keying order details into our system (we use Unleashed for inventory and Xero for invoicing). It's slow, error prone, and it doesn't scale. We've had orders missed entirely because they were buried in a long email thread. We want an AI-powered system that can: Monitor a shared inbox and classify each incoming email by intent (new order, reorder, quote, delivery status, complaint, other) Extract structured order data from both email body text and attached PDF purchase orders (product names or SKUs, quantities, delivery address, requested date) Match extracted products against our existing product catalogue, flagging anything ambiguous for human review rather than guessing Draft a reply for the staff member to approve, or auto-respond for simple cases like delivery status Push confirmed orders into Unleashed via its API What matters to us: accuracy over full automation. We would rather the system flag something for a human than create a wrong order. We need a clear confidence threshold and a human-in-the-loop review step. We're not technical, so we need someone who can recommend the right approach, not just build whatever we ask for. Tell us where you'd use an LLM versus rules, how you'd handle messy real-world PDFs, and how you'd stop it making things up. To apply, please tell us: A similar system you've built and what the outcome was How you'd approach the product matching and the hallucination risk Rough estimate of timeline and any assumptions you're making Any questions you have about our setup