Alvys wants to stop making you leave its platform to automate freight work. The company just launched Alvys Foundry, an agentic AI layer baked directly into its transportation management system. More than 20 pre-built agent templates cover the unglamorous but expensive corners of logistics: detention filing, document reading, rate audits, shipment tracking, asset compliance, and claims management. Customers can also upload an existing SOP or describe a task in plain language, and Foundry will generate a workflow for approval before anything touches live freight.
The pitch from CEO Nick Darman is that prior AI adoption in freight created tool sprawl — new logins, separate integrations, workflows that didn’t match how carriers actually operate. Foundry sidesteps that by running on the same infrastructure as Alvys’ TMS, which already has over 120 integrations and native EDI connections with hundreds of shippers. That means agents get real operating context: lane history, customer rules, margins, appointments. “We have the freight context and we understand your lanes,” Darman said.
Foundry’s governance layer, called Agent Shield, lets operators set approval and spending thresholds per agent. Higher-stakes actions pause for human sign-off, and every decision, action, and manual override lands in an audit trail. There’s also a model-routing system that shifts tasks between LLMs based on cost, speed, and quality — a nod to the reality that tying workflows to a single model provider gets expensive fast. Alvys says customer data won’t train public models, and the platform runs on a SOC 2-compliant foundation.
Alvys isn’t alone here. C.H. Robinson says its 30-plus agents have completed millions of tasks previously done by hand, including one that processes over 10,000 emailed pricing requests daily. Uber Freight has embedded more than 30 agents across procurement, tracking, and payments. The difference with Alvys is architectural: instead of layering automation on top of a TMS, it’s building execution directly into the system of record. Whether that makes agents more reliable or just harder to rip out if they misfire is the question nobody’s answered yet.
