AI Enabled Automotive Export Used Vehicle Brokerage

Client is a Japan-based used-vehicle export brokerage that helps international buyers — individuals and importer companies — purchase vehicles sourced from Japanese auto auctions, third-party sellers, and its own inventory. Customers are based in destination markets such as the UAE and Kenya. The business model centers on three coordinated roles: a customer submits a bid or inquiry, an agent manages the relationship and places the winning bid at auction, and an accountant handles invoicing, payment confirmation, and financial tracking, with every transaction crossing Japanese Yen and US Dollars.

Before the platform, Client ran its brokerage operation through manual, disconnected channels. Every bid request, invoice, and shipment update depended on someone remembering to follow up over chat or email, with no single system tying a customer's inquiry to the vehicle, the invoice, the payment, and the shipment that followed from it.

• Give customers a self-service platform to browse vehicles and submit bid requests or inquiries directly
• Track every bid and inquiry through a clear, auditable status pipeline — from submission to payment to shipment
• Give agents and accountants dedicated, role-based tools instead of routing everything through informal channels
• Handle every financial transaction in both JPY and USD with a consistent, auditable exchange rate

Success criteria

We built Client as a Laravel-based platform pairing a customer-facing application with a role-based Filament admin panel, so every participant — customer, agent, and accountant — works from the same underlying data. At its core is a set of explicit status pipelines: bid requests move from submission through review, auction placement, and a won/lost outcome; vehicle inquiries move from a quote through confirmation, invoicing, payment, and shipping.

Key capabilities

The solution was designed using a scalable modular monolith, API-first architecture, so the customer-facing SPA, admin panel, and real-time chat all consume the same underlying domain logic.

The implementation replaced ad hoc, manual coordination with a structured, auditable system across every stage of the customer journey.

• Unifying two customer funnels — auction bid requests and direct vehicle inquiries — under one invoicing and accountant-assignment system without duplicating logic
• Keeping dual-currency amounts consistent through a long-running workflow, where the exchange rate must stay fixed to the original transaction even as the live rate changes
• Automating status transitions safely, so that confirming an invoice advances its linked bid or inquiry without creating conflicting manual overrides

Each workflow — bid requests, vehicle inquiries, and invoices — was modeled as its own explicit state machine with a single source of truth for its allowed transitions. Cross-cutting effects, like an invoice confirmation advancing a linked inquiry to Paid, were implemented as one-way triggers from the transaction record to the workflow record, never the reverse — so there's always exactly one place that decides what happens next.

• Role-based access control (Spatie Laravel Permission + Filament Shield)
• Separate authentication guards for customers (Sanctum tokens) and admin staff (session-based)
• Signed, time-limited URLs for chat and invoice attachments, instead of permanently public files
• No payment gateway or card data ever touches the platform — all payments happen outside the system, with proof uploaded for manual accountant confirmation
• Role-scoped admin resources — agents cannot see invoice financials, accountants cannot see bid-approval controls
• Customer approval gating — new accounts must be email-verified and admin-approved before submitting a bid
• No PCI-DSS scope, since the platform never processes card payments directly

• Insurance quotes, status tracking, and documentation for shipped vehicles
• Scheduled, automated syncing of auction vehicle data from the external auction provider
• A hybrid mobile app for customers, already scoped as its own milestone
• Deeper integration with freight and insurance providers for real-time quotes
• Extending the existing admin dashboards with exportable reports
• Support for additional destination markets and multi-language customer-facing pages
• Expanding AuctionBot with additional skills beyond search, pricing, and comparisons

• Deep experience turning multi-party financial workflows into clear, testable state machines
• Familiarity with cross-border trade operations — multi-currency accounting, financial hand-offs, and role-based back offices
• Strong Laravel/Filament engineering, from real-time WebSocket messaging to PDF generation and third-party API integration
• A security-conscious approach to role separation and access control from day one
• Milestone-based delivery, with each phase shipped and validated before the next began
• Applied AI expertise — built AuctionBot, an OpenAI-powered RAG assistant with function-calling over live auction data

Client needed to turn a manual, chat-and-spreadsheet-driven brokerage operation into a system that tracks every bid, invoice, payment, and shipment in one place, across two currencies and three distinct roles. The platform replaces that manual coordination with an explicit, auditable workflow: customers submit bids and inquiries directly, agents and accountants each get the tools their role needs, and financial and shipping status advance automatically as the underlying work gets done.