Client handled a high volume of freight inquiries through multiple email inboxes, resulting in manual processing delays, inconsistent responses, and missed business opportunities. smartData developed Auto-Pilot, an AI-powered inquiry management platform that automatically captures inbound emails, classifies them into Type A (formal quotations) and Type D (rate requests), extracts key logistics information, and generates professional responses with minimal human intervention. Built using Python/Django, React, PostgreSQL, Gemini AI, and GCP, the platform provides centralized inquiry tracking, automated follow-ups, role-based approvals, and real-time visibility across the entire inquiry lifecycle, enabling faster decision-making and improved operational efficiency.

Features

• AI Inquiry Classification : Automatically reads and classifies unstructured inbound emails into Type A or Type D using Gemini-based NLP
• Automated Quotation Generation : Generates formal quotations with lane rates, validity, and T&Cs for Type A inquiries and sends via email automatically
• One-Click Rate Reply (Type D) : Drafts concise rate reply emails for simple inquiries with auto-approve or one-click send for pre-cleared clients
• Centralized Inquiry Dashboard : Role-based operational console tracking all inquiries with real-time status, approver actions, and follow-up reminders
• Follow-Up Automation & Closure : Sends scheduled follow-ups via email and auto-updates inquiry status through Won / Lost / No Response states
•  

Technical Challenges

• Parsing highly unstructured free-text emails with varying formats, languages, and attachment types (PDF, Excel, inline tables) to reliably extract logistics fields like origin, destination, vehicle type, and cargo specs. : Implemented Gemini-based NLP extraction with field-level confidence scoring and a fallback UI prompt that surfaces missing or ambiguous fields to a human operator for correction before the workflow proceeds.
• Building a reliable multi-mailbox connector that handles IMAP/SMTP connections across different email providers, manages deduplication of threaded replies, and avoids re-processing already-captured inquiries. : Designed a mailbox session manager with per-account UID tracking and idempotent inquiry creation logic, ensuring each email is captured exactly once regardless of polling frequency or network interruptions.
• Designing a dual-workflow system (Type A and Type D) that shares a common inquiry capture and tracking layer but diverges significantly in processing logic, output format, and approval rules without creating code duplication. : Applied a strategy pattern in the backend workflow engine – a common InquiryProcessor base class with separate TypeAHandler and TypeDHandler implementations, keeping shared logic (capture, status updates, follow-ups) in one place while allowing independent evolution of each path.
•