Freight software is getting a little less passive.
Alvys has launched Foundry, a new agentic AI platform built directly into its transportation management system, giving freight carriers and brokers AI agents that can actually carry out operational work rather than simply surface information or generate answers.
The distinction matters. A lot of logistics software already has AI somewhere in the product. Foundry is aimed at the repetitive work people still have to chase down after the software gives them the information.
That can mean filing detention, checking documents, following shipment updates, auditing rates or opening a claim.
And Alvys wants those jobs running inside the TMS itself.
More Than 20 AI Agent Templates Are Already Available
Foundry launches with more than 20 pre-built agent templates covering common freight workflows.
Among them is a Detention Agent, which can recognize when detention time has passed the permitted threshold and begin the filing process. A Track & Trace Agent can handle shipment status updates and check calls, while Alvys’ Document Intelligence can process freight paperwork such as rate confirmations, bills of lading and proof-of-delivery documents.
There are agents for rate audits, asset compliance and claims management as well.
The more interesting part may be what happens when a company needs something Alvys hasn’t already turned into a template.
Operators can upload an existing standard operating procedure or describe the job they want automated in plain language. Foundry then generates a workflow that can be reviewed before it goes live. Companies can also work directly with Alvys engineers on custom agents.
So this isn’t exactly a “tell the AI to run the company” setup.
There are still controls.
Freight Companies Can Test Agents Before Letting Them Touch Live Work
Alvys says agents can be tested against simulated data before being deployed into live freight operations.
Once running, operators can monitor or pause them from the Foundry platform. That is important in logistics, where a seemingly small automated decision can affect an invoice, driver, customer commitment or shipment already moving down the road.
Foundry also includes a governance layer called Agent Shield.
Companies can define approval requirements and spending limits for individual agents. Higher-impact actions can be held for human approval, and the system records agent actions, decisions and manual overrides in an audit trail.
That approach mirrors a wider enterprise trend toward combining automation with human oversight and common AI governance layers, particularly when agents begin touching consequential business processes.
In other words, Alvys is pushing autonomy, but not pretending oversight suddenly became optional.
The AI Already Knows the Freight Operation
One of Alvys’ bigger arguments for Foundry is that the agents aren’t arriving as another disconnected AI tool.
They’re sitting on top of the data already inside Alvys.
The company’s TMS includes more than 120 integrations along with native electronic data interchange capabilities. That gives its AI access to operational context such as customer rules, lane history, documents, appointments, margins and exceptions.
Alvys CEO Nick Darman said the company had watched customers experiment with AI products that introduced more software, more logins and more integration work instead of simplifying the operation.
Foundry goes in the opposite direction: put the agents where the freight data already lives.
Alvys describes the broader approach as AI that is “built in, not bolted on.” Its existing Alvys Intelligence offering already handles functions such as automatic load creation, real-time insights and agent-run workflows inside the TMS.
That may end up being more important than whichever model is underneath it.
Foundry Isn’t Locked to One AI Model
Alvys is also avoiding a hard commitment to a single large language model provider.
Foundry includes a model-selection system capable of routing different tasks between models based on considerations such as speed, quality and cost.
It’s a practical choice. The best model for interpreting an awkward freight document may not be the cheapest model for a routine workflow running thousands of times.
That same multi-model approach is appearing elsewhere in enterprise AI, including NVIDIA’s work on routing agent tasks between different AI models based on performance, latency and cost.
It also gives Alvys some room to swap models as the AI market changes instead of rebuilding its agent platform around whichever provider happens to be leading today.
On the data side, the company says Foundry operates on its SOC 2-compliant security foundation and that agreements with model providers prevent customer information from being used to train public models.
Freight Is Becoming an Early Test Ground for Agentic AI
Alvys isn’t alone here.
Freight may actually be one of the more natural industries for AI agents because the work is packed with repetitive, rule-driven processes that still involve emails, documents, status checks and humans moving information between systems.
C.H. Robinson has already deployed more than 30 AI agents for freight operations, according to reporting cited by Artificial Intelligence News. One of its agents reportedly processes more than 10,000 emailed pricing requests per day, while another converts information from load tenders and attachments into orders.
Uber Freight has taken a similar route. The company said in 2025 that it had more than 30 agents automating tasks across areas including procurement, tracking, shipment execution, payments and analytics.
The pattern is becoming hard to miss. The TMS used to be the place where freight activity was recorded. Now logistics companies want it to do some of the work.
Alvys Is Rolling Foundry Out Gradually
Foundry isn’t being opened to everyone at once. Alvys is introducing the platform through customer cohorts. The first cohort reportedly filled after the company demonstrated Foundry during a customer advisory board meeting in June, with a waitlist opened for another group.
Spartan Carrier Group is among the early users. Alvys quotes Spartan CEO Carlos M. Llanes Jr. saying the technology has allowed one person to handle work that previously required several people, while the company positions the agents as a way to move employees away from repetitive tasks and toward work requiring judgment and customer service.
The launch also comes less than a year after Alvys raised a $40 million Series B round in September 2025. AI News reports the company has raised $77 million overall and says Alvys now handles more than $9 billion in freight through its platform annually.
AI Agents Are Moving From Demo Material Into Actual Operations
Agentic AI has spent plenty of time looking impressive in demonstrations. Freight gives it a tougher assignment. Documents aren’t always clean. Exceptions happen constantly. Customers have different rules. Loads change. Drivers are moving. Money is involved.
An AI agent that can reliably operate in that environment is considerably more useful than another chatbot sitting beside the TMS.
That same shift from demonstrations toward operational deployment is showing up across other industries, including enterprise AI systems being integrated directly into finance, procurement and customer operations.
That’s what makes the Alvys launch worth watching. Foundry isn’t trying to replace the transportation management system with AI. It’s turning the existing system into something that can take action. For freight operators buried in repetitive administrative work, that’s a much more concrete pitch.
