Shipsy launches AI intelligence layer for enterprise logistics

AI-native enterprise logistics management platform Shipsy has launched the beta version of Shipsy Brain, a logistics intelligence layer designed to help enterprises make decisions and execute operational workflows with greater accuracy, speed and control.

Shipsy Brain combines specialised open-source models with logistics data generated through the Shipsy platform. It is designed to help enterprise AI agents understand logistics-specific context and move beyond dashboards and copilots to systems that can reason, recommend and act within defined business controls.

“AI in logistics must understand how shipments, drivers, documents, carriers, contracts and many other variables interact with each other and then take the right action. Shipsy Brain brings this operational depth to global supply chains,” said Soham Chokshi, Co-founder and CEO of Shipsy.

The intelligence layer is built on more than five years of logistics data from the platform. This includes more than 50 billion operational events, over 1 billion auto-assigned decisions and 500 million human decisions, along with reassignment history. It also includes more than 100 billion GPS location pings linked to route and delivery outcomes, over 3 billion delivery labels across carriers and formats, 342 carrier integrations, more than 50 million routing decisions and over 3 billion hub scans.

Shipsy Brain also uses more than 5,000 workflows built across Shipsy’s Workflow Builder and AgentFlow platform. The data allows it to understand logistics-specific context that may be ambiguous to general AI models.

The system acts as a central intelligence layer that coordinates specialised models for areas including documents, consignments, trips, workflows and finance. These models can support agents handling document validation, address intelligence, anomaly detection, ETA prediction, routing, settlement management and workflow recommendations.

Shipsy Brain operates within Shipsy’s AgentFleet platform. It monitors live operations, identifies manual work that can be automated, requests permission and executes approved actions based on its confidence in each action. When a human corrects the system, it learns from the correction.

As a context layer, Shipsy Brain is also model agnostic. Shipsy said any model connected to the layer can access the full operational context, allowing improvements in general AI models to increase the usefulness of the system.

Shipsy said the system is designed to improve accuracy by using proprietary knowledge from more than 5 billion shipments, billions of platform actions and thousands of logistics workflows. This helps it interpret logistics-specific details, such as different names used for consignment numbers.

In document-intelligence benchmarks, the Shipsy model recorded an overall field extraction score of 82.2%, compared with 63.4% for Gemini 3.5, 62% for Gemini 3 and 62.4% for Gemini Pro. Its logistics domain knowledge score was 92.4%, compared with 45.9%, 38.6% and 45.9%, respectively.

For document references, Shipsy Brain scored 83.6%, compared with 51.6% for Gemini 3.5, 52.8% for Gemini 3 and 44.1% for Gemini Pro. Its document understanding score was 86.6%, compared with 81.2%, 82.7% and 78.9%, respectively.

Shipsy said the use of fine-tuned logistics-specific models can also improve speed because the models already understand industry terminology, workflows and operational context. This reduces the need for extensive prompts required by general-purpose models and can accelerate decisions, customer support and the deployment of logistics use cases.

The company said Shipsy Brain is also designed to reduce costs by using fine-tuned, self-hosted open-source models instead of relying entirely on expensive frontier-model tokens. Specialised models require less computation and fewer tokens for logistics tasks, while giving enterprises greater predictability and control over costs.

The gap does not remain steady and widens over time. Every shipment, decision and human correction feeds back into Shipsy Brain, allowing it to learn from corrections and improve based on how it is used.

Shipsy Brain is currently available in beta for selected enterprises.