Alvys 在其货运 TMS 中加入了一个可自建的 AI 代理层

Why it matters

Freight AI is moving from standalone bots into systems of record. Alvys is betting that operator-controlled logic, approvals and data access can bring agents to smaller fleets while making its TMS much harder to replace.

Nick Darman, founder and CEO of Alvys, is turning the transportation management system he originally built for his own freight operation into a place where carriers and brokers can build AI workers. Alvys 在8月17日宣布 that Alvys Foundry will let customers deploy prebuilt agents, assemble their own in plain English or have Alvys engineers create them.

The launch follows a familiar Darman pattern: take a problem he encountered as an operator, build the missing software and then sell the resulting system to the rest of the freight market. Darman grew up in a trucking family, started the asset-based brokerage Archerhub in 2014 and assembled an engineering group after existing TMS products failed to fit its workflows. Alvys says the internal system helped Archerhub pass $90 million in annual revenue before Darman turned it into a standalone software product in 2020.

Darman paired that freight experience with co-founder and CTO Leo Gorodinski, Jet.com's first employee and former vice president of engineering. Walmart acquired Jet.com for $3.3 billion in 2016. That division of labor has shaped Alvys from the beginning: Darman supplies the operating scars, while Gorodinski brings experience building large cloud systems.

Foundry is the clearest attempt yet to package that combination as a platform rather than a collection of AI features. Alvys already offered AI-assisted load creation, a reporting assistant called Insights and real-time weather and load alerts through Intel. Foundry adds an agent layer that can carry out multi-step operating procedures using the load, customer, document and financial data already stored in Alvys.

A controlled launch, not a universal switch

Alvys says Foundry includes more than 20 templates covering detention filing, document intake, rate auditing, track-and-trace, claims, compliance and load creation. Users can describe a workflow in plain English, review the generated process, test it against simulated data and then deploy it on a selected lane or a larger book of freight.

The second path is conventional: choose a prebuilt agent for a standard workflow. The third puts Alvys engineers into the implementation, building and tuning an agent for a customer that lacks its own technical staff.

The product is beginning with a controlled rollout. Alvys previewed Foundry at its 6月17日 customer advisory board meeting, filled the initial customer cohort and opened a waitlist for a second group. That sequence matters because autonomous actions inside a TMS can touch invoices, compliance records, customer communications and payments. Foundry is being introduced as operating infrastructure, though the initial deployment is closer to a staged production test than a broad release to every Alvys user.

Alvys is pairing the agents with a governance layer called Agent Shield. According to Alvys, administrators can set spending and approval thresholds, require human sign-off for consequential actions, pause agents and inspect an audit trail of decisions and overrides. Alvys also says enterprise agreements prevent model providers from training on customer data.

The architecture includes a model router that assigns work based on cost, speed and quality rather than committing every task to one large language model. The approach could become important once an agent moves from a demonstration to processing thousands of loads. A premium reasoning model may make sense for an unusual claims dispute. Reading a standard proof-of-delivery document may call for something cheaper.

The TMS is becoming the distribution channel

Alvys is entering a freight software market that has moved quickly from AI assistants to agents embedded in the system of record.

Trimble announced AI agents for order intake, invoice scanning and roadside breakdowns in November 2025. Tai TMS introduced an automated track-and-trace agent in April 2026. Transfix previewed AI exception management in May. On August 5th, McLeod Software and Augment announced that Augment's agent would operate inside McLeod's PowerBroker TMS for carrier selection and shipment tracking.

That competitive activity narrows the significance of Alvys' claim that Foundry is built in rather than bolted on. Embedded AI is rapidly becoming table stakes in transportation management. Alvys' sharper distinction is the attempt to give operators a builder, reusable templates and implementation help within the same product.

Darman is framing that design as an answer to vendor dependence. In the launch announcement, he recalled paying software providers for each customization, report and integration while running his freight operation. "Vendors got me started, then charged for every customization, every report, every integration," he said.

Foundry turns that frustration into product strategy. Alvys wants customers to encode standard operating procedures themselves instead of sending every change request to an automation vendor. That is attractive in freight, where two carriers moving similar loads can still have different escalation rules, customer scorecards, appointment practices and approval limits.

The promise also creates a new form of dependence. Once a freight operator has converted its procedures into agents that run on Alvys data and integrations, changing the underlying TMS becomes a larger migration. The customer may control the workflow blocks, but Alvys owns the environment where those blocks access operational context and take action. Foundry can give operators more control over individual automations while making Alvys harder to replace.

That is sound vertical-software strategy. AI features are easy to demonstrate and increasingly easy for competitors to copy. Deeply encoded operating procedures, approvals, integrations and historical freight data create a stronger retention mechanism.

Darman's $40 million AI commitment reaches the product

Foundry is the first major test of the AI plan Alvys attached to its 2025年9月29日的 Series B. RTP Global led the $40 million round, with Alpha Square Group, Titanium Ventures, Picus Capital and Bonfire Ventures participating. Alvys says it has raised $77 million in total.

At the time, Darman said the capital would put AI at the center of dispatch, decision-making and back-office workflows. Foundry moves that spending from a roadmap statement into a product that customers can test. It also gives Alvys a route to expand beyond TMS subscriptions by becoming the automation layer for work that previously required staff, outsourced services or separate point products.

Alvys has not described Foundry's commercial model in the launch materials. Its economics will still shape adoption. Model calls, implementation help and continuous agent execution introduce variable costs that ordinary seat-based software does not carry. The intelligent model router addresses the compute side, while templates can reduce the amount of engineering Alvys must provide for each customer.

The other part of the expansion is customer size. Alvys said on August 17th that its TMS would be accessible to fleets of all sizes, including owner-operators and small carriers. That returns Alvys to Darman's original thesis: smaller freight operators should have access to software capabilities historically reserved for larger carriers and brokers.

Alvys 报告称其拥有超过 2,000 家客户公司、超过 3,000 个机动车承运人许可、150 多名员工，以及通过其软件每年处理超过 70 亿美元的货运量。这些数据来自 Alvys，本公司尚未披露收入和 Foundry 的采纳情况。不过，已安装的客户基础为 Alvys 提供了独立货运代理供应商必须通过整合和合作关系构建的东西：对决定代理可执行操作的记录和工作流的直接访问权限。

真正的考验是在生产规模上完成枯燥的工作

Foundry 最初最有力的用例刻意选择了不那么光鲜的场景。滞留跟踪、文件分类、状态查询电话和发票准备是经常发生的工作，遵循可重复的规则，消耗人力，通常不需要战略判断。它们也会留下管理者可以审计的清晰记录。

这使得它们成为谨慎运营商的实际切入点。承运人可以先从一个能够识别迟到预约、计算滞留何时开始并准备所需通知的代理开始。人工可以在通知到达客户之前批准该通知。运营方在不将定价、安全或高价值付款控制权交给代理的情况下，获得可量化的时间节省。

Alvys 的客户 Spartan Carrier Group 正作为启动参考。创始人兼 CEO Carlos M. Llanes Jr. 是一位美军作战老兵，他在 Fort Worth 围绕准时制和汽车货运建立了这家承运人，他将 Foundry 描述为一种嵌入式研发资源。他的背书支持了 Darman 的论点，即较小的货运运营商可以在不雇佣内部 AI 团队的情况下获得技术能力。这仍然是定性描述，而非生产基准。

Alvys 需要证明普通运营人员在实施专家离场后能够维护这些代理。以普通英语构建的工具可以简化工作流的初稿。货运流程仍然包含边缘情况、客户例外和遗留习惯，这些很少能符合一个整洁的演示。Foundry 的价值将取决于运营商识别故障、理解故障原因并在不再生成新的支持工单的情况下调整工作流的速度。

Darman 多年来一直主张货运软件应当符合运营者现有的工作方式。Foundry 将这一论点推进到 AI 领域：保留程序、暴露控制并自动化围绕它的重复执行。如果 Alvys 能让该闭环可靠运行，TMS 就不再是员工在完成工作后更新的数据库。它将成为大量工作得以完成的地方。