Build a Custom AI Dispatcher for Your Freight Brokerage

Most freight brokerages run on a Transportation Management System (TMS) like McLeod, TMW, or MercuryGate. These systems are excellent for record-keeping but their built-in carrier matching is rudimentary. They can filter for carriers who have run a lane before but cannot rank them by predicted profitability, on-time performance for that specific shipper, or current driver availability. The logic is based on static rules, not learned patterns from your own historical data.

This forces a huge amount of manual work. Consider a 15-person brokerage where dispatchers receive hundreds of tender emails daily. A dispatcher spends 10 minutes reading each email, manually keying origin, destination, weight, and equipment type into the TMS. A single typo in a ZIP code can send the entire process sideways. After data entry, they begin the slow process of calling or emailing carriers one-by-one, working from a simple filtered list.

The structural issue is that TMS platforms are designed as databases, not as intelligent decision engines. Their architecture is optimized for storing and retrieving structured data, not for processing unstructured emails or running complex ranking algorithms. They cannot easily incorporate your team's 'tribal knowledge'—like knowing which carriers are best for multi-stop loads or which ones are most reliable on a Friday afternoon. You are stuck with the vendor's rigid data model and workflow.

The first step is always a data and workflow audit. Syntora would connect to your TMS to pull 12 months of historical load data and analyze a sample of 200-300 tender emails from your top shippers. This audit identifies the key data points for matching, establishes a baseline for data quality, and maps exactly how a new system would fit into your dispatchers' current workflow. You receive a technical scope document outlining the integration points and data models.

We would build a Python-based service using the Claude API to parse unstructured text from emails into clean, structured JSON. We have built similar document processing pipelines for financial services, and the same pattern applies directly to logistics. This structured data then feeds a FastAPI service running on AWS Lambda. The service queries your carrier data from a Supabase database, ranks the top 5 candidates based on your custom logic, and pushes the recommendations back into your TMS via its API.

The delivered system augments your existing TMS; it does not replace it. Your dispatchers would see a new panel in their current software showing the automatically parsed load data and a ranked list of carrier recommendations. They make the final decision. You receive the complete source code, a runbook for maintenance and monitoring, and direct ownership of the system running in your own cloud account.