Freight software company Alvys is bringing configurable AI agents directly into its transportation management system. The new platform, Alvys Foundry, is designed to automate the repetitive operational work that surrounds moving goods, while keeping consequential decisions subject to human approval.
The pitch is broader than adding a chatbot to a logistics dashboard. Foundry is intended to turn standard operating procedures into running workflows that can handle documents, tracking, detention, rate audits, asset compliance and claims. Customers can begin with more than 20 prepared templates, upload an existing procedure or describe a process in ordinary language.
From written procedures to active freight workflows
Alvys says a workflow can be tested with simulated data before deployment, then monitored or paused once it is live. That sequence matters in freight, where a seemingly minor automation error can affect a shipment, a payment or a customer relationship.
The agents operate with the context already held in the Alvys TMS. The company points to more than 120 integrations and native electronic data interchange connections with hundreds of shippers. In principle, that gives an agent access to the operational signals required to complete a task without forcing staff to move information between separate systems.
The flexibility could be particularly useful for carriers and brokers whose processes do not fit a universal template. A team might automate a routine document check while leaving an unusual claim or costly exception for an employee. The source does not provide measured results across a large customer base, however, so the effect on processing time, error rates and staffing remains to be demonstrated.
Guardrails are part of the product, not an afterthought
Foundry includes what Alvys calls Agent Shield. Administrators can define approval and spending thresholds, require human sign-off for high-impact actions and retain an audit trail. Those controls are important because a workflow that can act inside a TMS carries more operational risk than an assistant that merely drafts an answer.
Alvys also says its environment is SOC 2 compliant and that provider agreements prevent customer information from being used to train public models. These are company claims rather than an independent security assessment presented in the source. Buyers will still need to examine which data each agent can reach, how exceptions are handled and whether the audit record is sufficiently detailed for their own compliance obligations.
The system can route work among models according to cost, speed and quality. That is meant to reduce dependence on a single model provider and avoid using an expensive system for every task. It also introduces a management challenge: performance can vary between models, so teams need to know whether a workflow behaves consistently when its underlying model changes.
Logistics becomes a proving ground for practical agents
Alvys is entering a field where large incumbents are already pursuing agentic automation. C.H. Robinson has described more than 30 agents handling millions of tasks, including a pricing agent processing over 10,000 emails a day. Uber Freight has also said it operates more than 30 agents. Those examples indicate that freight contains abundant structured, repetitive work, although the companies’ figures are not directly comparable.
Early Alvys customer Spartan Carrier Group is part of the initial rollout. The first cohort is filled, with a waitlist open for a second. That limited launch should provide evidence about where customer-built agents work reliably and where human intervention remains essential.
The company raised a $40 million Series B in September 2025, bringing total funding to $77 million, and says its platform handles more than $9 billion in freight annually. Those figures give Foundry a meaningful operational setting. They do not guarantee that natural-language workflow creation will be dependable at scale.
Foundry’s most consequential idea is that agents should live inside the system of record, surrounded by permissions and approvals. If Alvys can make that design manageable for ordinary freight teams, agentic AI may shift from isolated experiments to routine operations. The staged rollout will show whether the controls remain as useful as the automation becomes more ambitious.
