AI Agents Are Starting to Run Freight Operations as Alvys Launches Foundry

 

Artificial intelligence is moving deeper into the logistics industry, and this time it is not just helping employees analyze information.

It is beginning to perform the work itself.

Freight software company Alvys has introduced Alvys Foundry, an agentic AI platform designed to automate operational tasks directly inside its transportation management system, or TMS.

Instead of requiring carriers and freight brokers to open a separate AI application, Foundry places AI agents inside the same platform where companies already manage shipments, documents, rates, customers, assets, and exceptions.

The platform launches with more than 20 pre-built AI agent templates, while customers can also create their own agents or work with Alvys engineers to configure workflows around their specific operating procedures.

From AI Assistant to AI Worker

Traditional generative AI systems are usually designed to answer questions, summarize information, generate documents, or recommend actions.

Agentic AI goes further.

An AI agent can be given a goal, follow a defined workflow, access approved systems and data, and perform actions on behalf of the user.

For freight companies, that could mean automatically checking whether a truck has exceeded free detention time, preparing the required documentation, updating shipment status, auditing an invoice, or opening a cargo claim.

That is the direction Alvys is taking with Foundry.

The platform's agents are designed to perform specific steps in transportation workflows rather than simply provide suggestions.

What the Alvys AI Agents Can Do

Among the agents included in Foundry is a Detention Agent, which can identify when allowable waiting time has been exceeded and initiate the detention process.

Another feature called Document Intelligence can read, identify, and file freight documents including:

• rate confirmations
• bills of lading
• proofs of delivery

A Track & Trace Agent can manage shipment check calls and status updates.

Alvys has also developed templates including a:

Rate Audit Agent, which examines freight invoices and rates.

Asset Compliance Agent, which checks information such as operating authority, insurance, and safety records.

Claims Agent, which helps open and document freight claims.

Instead of manually programming every automation, operators can upload an existing standard operating procedure, or SOP, or simply describe the task they want automated using natural language.

Foundry can then generate a proposed workflow for the company to review and approve.

That could significantly lower the technical barrier to deploying AI automation.

Test the AI Before Giving It Real Freight

Allowing an AI agent to take actions inside an operational system introduces obvious risks.

A wrong chatbot answer is inconvenient.

A wrong AI action involving a shipment, customer, payment, claim, or freight rate could become expensive.

Alvys says Foundry allows companies to test agents using simulated data before deploying them into live operations.

Operators can then build, deploy, monitor, and pause agents through the Foundry environment.

This gives companies an opportunity to validate how an agent behaves before granting it access to real freight workflows.

Why Put AI Inside the TMS?

According to Alvys CEO Nick Darman, one of the problems the company observed in early enterprise AI adoption was fragmentation.

Companies were adding AI applications on top of their existing software.

That sometimes meant additional interfaces, separate logins, new integrations, and workflows that employees had to learn.

Foundry takes a different approach.

Rather than building another layer outside the transportation platform, Alvys is placing AI agents directly into its existing TMS.

That matters because the system already understands much of the company's operational environment.

Depending on the customer's configuration, that could include:

• historical shipping lanes
• customer rules
• freight documents
• appointments
• margins
• shipment exceptions
• carrier information
• operating procedures

This contextual information can make an AI agent considerably more useful than a generic chatbot that receives only a single prompt.

As Darman explained, Alvys already has access to the freight context and understands the lanes its customers operate.

More Than 120 Integrations

Alvys says its TMS infrastructure includes more than 120 integrations, together with native electronic data interchange connections involving hundreds of shippers.

Those connections potentially give Foundry agents access to the same operational ecosystem already used by freight companies.

This is important because enterprise AI agents are only as useful as the systems they can securely access.

An agent that understands what needs to happen but cannot interact with the transportation software, documents, pricing systems, or shipment data still requires a human to complete the process.

By embedding AI directly inside the TMS, Alvys is attempting to close that gap.

Human Approval Is Still Part of the System

Alvys is also adding governance controls through a feature called Agent Shield.

Companies can configure approval requirements and spending thresholds for individual AI agents.

Lower-risk actions could potentially proceed automatically, while more consequential decisions can be stopped until a human approves them.

The system also keeps an audit trail of:

• agent decisions
• actions performed
• human interventions
• manual overrides

This becomes increasingly important as AI systems move from generating recommendations to executing business transactions.

If something goes wrong, companies need to know what the agent did, why an action occurred, and whether a human approved or modified the decision.

The AI Does Not Have to Use Just One Model

Foundry also includes a system for choosing between different large language models.

Instead of permanently connecting every workflow to one AI provider, the platform can route tasks depending on factors such as:

cost, speed, and output quality.

That could allow a company to use a powerful but more expensive model for complicated reasoning while sending simpler tasks to a faster or cheaper model.

This multi-model strategy could become increasingly common in enterprise AI.

Businesses may ultimately care less about which AI brand is underneath their software and more about whether the system completes the task accurately, quickly, securely, and at an acceptable cost.

Freight Companies Are Already Deploying AI Agents

Alvys is not alone.

The freight industry is emerging as one of the clearest enterprise use cases for agentic AI because logistics contains enormous numbers of repetitive but structured operational tasks.

C.H. Robinson has already deployed dozens of AI agents across freight operations.

The company previously reported that more than 30 agents had collectively handled millions of tasks that were previously performed manually.

One of its AI agents reportedly processes more than 10,000 emailed pricing requests every day.

The system can read the request, retrieve a freight price from the company's pricing infrastructure, and send a response.

Another agent can read load tenders and attachments before turning the information into transportation orders.

C.H. Robinson Vice President of Artificial Intelligence Mark Albrecht described the distinction clearly.

Agentic AI does not simply analyze information or create content.

It can take actions toward completing a defined objective.

Uber Freight Is Moving in the Same Direction

Uber Freight has also incorporated AI agents into its transportation management technology.

In May 2025, the company said it had more than 30 AI agents automating work involving:

• freight procurement
• shipment execution
• tracking
• payments
• analytics

Its longer-term strategy reflects a broader change happening across enterprise software.

Historically, a transportation management system primarily functioned as a system of record.

It stored information about shipments and helped workers manage transportation operations.

With AI agents, that model begins to change.

The software can potentially become a system of action.

Instead of merely displaying that a shipment is late, the system could investigate the problem, contact the appropriate party, update the records, recommend a response, or execute an approved workflow.

A Bigger Shift in Enterprise Software

This transition extends far beyond logistics.

For decades, business software depended on humans navigating menus, completing forms, moving information between systems, checking dashboards, and following standard procedures.

Agentic AI introduces another possibility.

Employees could increasingly describe the desired result while AI systems perform portions of the workflow.

That does not necessarily eliminate the human worker.

It changes where human effort is concentrated.

Employees could spend less time copying data, checking routine documents, sending repetitive updates, or processing standard exceptions.

Human workers could instead focus on negotiation, customer relationships, unusual problems, judgment calls, strategy, and decisions where context matters.

That appears to be the approach being taken by Spartan Carrier Group, one of the early Foundry customers.

Founder and CEO Carlos M. Llanes Jr. said the platform is being used to reduce manual freight work while allowing employees to remain focused on judgment and customer service.

Security Becomes More Important When AI Can Act

The more authority AI agents receive, the more important enterprise security becomes.

According to Alvys, Foundry operates on a SOC 2-compliant security foundation.

The company also says its agreements with AI model providers prevent customer information from being used to train publicly available models.

Those safeguards are significant because transportation systems can contain commercially sensitive information including pricing, customer relationships, shipment details, operating margins, and carrier data.

Giving AI access to those systems creates new productivity opportunities.

It also expands the importance of access controls, audit logs, permission management, data protection, and human oversight.

Alvys Continues Its Expansion

The Foundry launch follows Alvys' $40 million Series B funding round in September 2025, led by RTP Global.

The company has reportedly raised $77 million in total funding.

Alvys also says its platform currently handles more than $9 billion worth of freight annually.

Foundry is not being released immediately to every customer.

Instead, Alvys is introducing the platform through controlled customer cohorts.

The first group reportedly filled after the company demonstrated Foundry during a customer advisory board meeting on June 17.

A waiting list has since opened for another cohort.

That phased deployment could also give Alvys more opportunity to observe how AI agents behave in real transportation environments before expanding availability.

Why This Matters

The biggest story here is not simply that another company has added AI to its software.

Companies have been doing that for years.

The important shift is what the AI is being allowed to do.

The first wave of generative AI helped workers create.

The next wave helped workers analyze.

Agentic AI is increasingly being designed to execute.

And logistics could become one of its biggest testing grounds.

Freight operations contain thousands of repetitive decisions and transactions involving documents, rates, schedules, tracking updates, compliance checks, claims, and payments.

Automating even part of that workload could materially change how transportation companies operate.

But it also raises a new question for every enterprise adopting AI:

How much authority should we give the machine?

The companies that answer that question successfully may not simply have AI assistants sitting beside their employees.

They may eventually have entire fleets of specialized AI agents working alongside them.