Use AI to Track Logistics Vendor Performance and Reduce Risk

Many logistics SMBs rely on their Transportation Management System (TMS), like MercuryGate or AscendTMS, for basic tracking. These platforms are effective for managing shipment lifecycles, but their vendor scorecards are rudimentary. A TMS can track on-time performance but cannot parse unstructured data from carrier emails or performance review PDFs. Adding custom metrics like 'damage claim resolution time' is often impossible without expensive add-on modules.

Consider a 15-person freight brokerage working with 30 different carriers. Every month, their operations manager spends two full days compiling data. This process involves logging into 15 separate carrier portals to download CSVs, manually extracting numbers from 10 others who email PDF summaries, and chasing the remaining 5 for data. The resulting Excel spreadsheet is inconsistent, filled with copy-paste errors, and is already two weeks out of date by the time it is finished.

The structural problem is that TMS platforms are built for transactional data, not analytical insight. Their data models are rigid, designed to track a load from origin to destination. They are not engineered to ingest, parse, and score unstructured text from contracts, emails, or PDF reports. To add this capability would require a complete architectural shift, so they instead rely on manual data entry.

As a result, contract negotiations are based on anecdotal evidence, not data. High-risk carriers remain in the network longer than they should because their poor performance is buried in disconnected systems. Identifying the top 5% of carriers for a new high-value lane becomes a guesswork exercise, not a data-driven decision.

The engagement would start with an audit of your top 10 to 15 carriers. Syntora would document how each one provides performance data: a documented API, CSV downloads from a portal, PDF reports via email, or unstructured email updates. This audit produces a data-source inventory and a technical plan. You receive a clear map of what data is available and how it will be unified into a single scoring model.

The core of the system would be a Python service running on AWS Lambda, which keeps hosting costs under $20 per month. For carriers with APIs, the service makes scheduled calls using `httpx`. For portals without APIs, it uses browser automation to log in and download reports. For PDFs and emails, the Claude API extracts and structures the data into a consistent format. We've used this exact Claude API pattern to process complex financial documents; the same technique applies directly to carrier performance reports. All structured data is then stored in a Supabase Postgres database.

The delivered system includes a dashboard and an API. The dashboard, built with Streamlit, provides a real-time leaderboard of all carriers scored against your key metrics (on-time delivery, tender acceptance rate, claim resolution time). A FastAPI endpoint allows your existing TMS to pull a carrier's score on demand, integrating the insight directly into your daily workflow. You own all the code, deployed in your AWS account.