Build a Custom AI Route Planner for Your Fleet

Many small trucking companies start by using Google Maps for routing. This works for a single truck going from point A to B, but it fails for optimizing a 15-truck fleet with 100 stops. A dispatcher spends hours manually grouping stops by zip code and guessing the best sequence, a process that cannot account for driver hours of service (HOS) or vehicle capacity constraints across the entire fleet.

The next step is often the route planning module included in a TMS like Motive or Samsara. These tools can sequence stops for a single driver, but they are not dynamic. Consider a dispatcher for an LTL carrier who gets an urgent pickup request at 10 AM. The TMS cannot identify which driver is best positioned to take the job based on their current location, route, and remaining capacity. The dispatcher must call multiple drivers, disrupt planned routes, and manually update the system, taking 30-45 minutes of valuable time.

Off-the-shelf routing software like Route4Me offers more advanced optimization but imposes a rigid, one-size-fits-all model. These platforms struggle with unique business rules. For instance, a food distribution company may require refrigerated trucks to return to base by 4 PM for washout. A standard routing tool optimizing for the lowest mileage will not respect this critical operational constraint, creating compliance issues.

The structural problem is that these tools are closed platforms. They are not designed to integrate with your specific business logic or other internal systems. A small trucking company's competitive advantage comes from its unique operational model, but generic software forces it into a generic workflow, leaving efficiency and money on the table.

The first step is a data and process audit. Syntora would analyze 3 months of your historical delivery data from your TMS to understand travel times, stop durations, and common delay patterns. We would map out every specific constraint your business has, from vehicle types and capacities to driver schedules and customer-specific delivery windows. This audit produces a clear plan of action and confirms the data is sufficient to build an effective model.

The technical core would be a Python service using an open-source optimization engine like VROOM, wrapped in a FastAPI endpoint for integration. This service ingests daily manifests from your TMS and enriches them with data from a real-time source like the TomTom Traffic API. For complex, unstructured delivery notes, we can use the Claude API to parse text and convert it into structured instructions for drivers. The entire system runs on AWS Lambda, so you only pay for compute time when routes are being calculated, keeping monthly costs under $50.

The delivered system plugs directly into your existing TMS. Your dispatchers use their familiar interface, but a new 'Optimize Routes' button sends the day's jobs to the AI engine and gets a fully optimized plan for the entire fleet back in under 30 seconds. A webhook is also exposed for dynamic re-routing. When a new job is created in the TMS, it triggers the service to find the best driver assignment instantly.