Automate Carrier Selection and Rate Negotiation with Custom AI

Small logistics companies often rely on their TMS, like AscendTMS or MercuryGate, for carrier management. These platforms are excellent for tracking loads but their rating engines are rigid. They work well with carriers who offer direct API integrations, but fail when a key partner sends rates in a weekly PDF email or has an archaic web portal. The TMS cannot read the PDF, forcing manual data entry.

Consider a 15-person 3PL that needs to quote a new lane from Dallas to Chicago. The TMS pulls rates from three major carriers via API. However, two of their most reliable regional carriers for that lane do not have APIs. One emails a 200-page PDF rate sheet every Monday, and the other requires logging into a web portal with a CAPTCHA. A dispatcher spends 15 minutes downloading the PDF, searching for the right lane, then logging into the portal, re-entering shipment details, and copy-pasting the two rates back into the TMS.

This manual process introduces two major risks. First, data entry errors are common, leading to under-quoted jobs that lose money or over-quoted jobs that lose the business. Second, it is slow. By the time the dispatcher has gathered all the rates, the best option from an API-connected carrier might have already been booked by a competitor. Dispatchers are forced to choose between being thorough and slow, or fast and incomplete.

The structural problem is that TMS platforms are built for standardized data exchange via APIs. Their architecture is not designed to handle the unstructured, inconsistent formats common with smaller, regional carriers. They offer a one-size-fits-all integration model, and if a carrier does not fit, the burden of manual work falls on your team.

The engagement would begin with an audit of your top 10-15 carriers. We would document how each provides rate data: API documentation, PDF rate sheets, Excel files via email, or web portals. This audit determines the right tool for each carrier and shapes the project plan. You would receive a clear breakdown of the integration strategy for each carrier before any code is written.

The core of the system would be a FastAPI service hosted on AWS Lambda. For carriers with APIs, it would use httpx for parallel, asynchronous calls. For carriers sending PDFs or emails, a Python script using the Claude API would parse documents and extract structured rate data. For web portals, Playwright would automate browser interactions. All data would be normalized into a single format and stored in a Supabase Postgres database for historical analysis.

The final deliverable is a simple API endpoint that your team can access via a web interface or integrate directly into your existing TMS. You would enter the load details once, and the system would return a ranked list of all available carriers and their rates in under 5 seconds. The system also includes the full Python source code, a runbook for maintenance, and a simple dashboard to monitor carrier response times and parsing accuracy.