Barrett plans network-wide warehouse automation layer

Barrett Distribution Centers plans to deploy UNIT AI’s Networked Physical AI Platform across its US logistics operation from 2027, extending warehouse automation from individual buildings into decisions covering inventory, fulfilment, and returns across the wider network.

The expanded partnership will introduce distributed inventory placement, fulfilment orchestration, network visibility, and decentralised returns processing. UNIT is positioning the system as a control layer capable of coordinating activity between warehouse locations rather than automating one isolated task inside a single building.

Barrett already uses UNIT robotics within its fulfilment operations. The next phase changes the scope of the project by connecting decisions across several logistics nodes, allowing inventory location and order allocation to be considered against conditions elsewhere in the network.

Most warehouse automation remains strongly tied to individual facilities. Automated storage reduces travel in one building, robots move stock between local processes, and sortation equipment increases dispatch capacity at a particular site. Those systems may be sophisticated, but the operating decisions surrounding them can still be made largely within the boundaries of each warehouse.

A network model introduces another layer. Inventory held in one location can potentially be considered against stock, demand, capacity, and returns activity in another, allowing a 3PL to decide which facility should process work rather than assuming the nearest or historically assigned warehouse remains the best option.

That has direct implications for inventory positioning. Replicating the same products across several facilities can shorten delivery distances but increases the quantity of stock required across the network. Concentrating inventory reduces duplication but can produce longer transport movements and greater dependence on particular sites.

Software capable of coordinating distributed inventory is intended to balance those factors more dynamically. The quality of the result will depend on whether the system has accurate information about available stock, orders, warehouse capacity, and the physical capability of the automation operating at each location.

Returns add another complexity. Ecommerce products are often routed towards a central return operation even when an item could be assessed and returned to saleable inventory closer to another customer. Barrett and UNIT plan to support decentralised return processing so stock can potentially re-enter the network nearer the point where it is needed.

The approach remains dependent on physical processes. Software may select a warehouse for a particular order, but that building still requires sufficient storage, robots, workstations, labour, packing capacity, and carrier collections to complete the work within the required service window.

That creates a risk of shifting constraints rather than eliminating them. Directing additional volume towards a facility with available inventory may improve stock utilisation but overload its packing area or outbound docks. Network orchestration therefore has to reflect the real operating limits of machinery and labour rather than treating buildings as unlimited nodes on a diagram.

Transportation also affects the calculation. A warehouse with spare picking capacity is not necessarily the cheapest location from which to fulfil an order if parcel distances, linehaul requirements, or carrier availability increase significantly. Decisions spanning several facilities need to include the physical freight network as well as warehouse activity.

UNIT is delivering the platform through a Warehouse-as-a-Service model. The companies say this will allow Barrett to add robotics incrementally without the level of upfront capital expenditure associated with some fixed automation projects.

The commercial model is relevant to a 3PL because customer contracts and product profiles change more frequently than the physical life of warehouse machinery. Equipment installed around one large customer can become difficult to justify if volumes fall or the account moves elsewhere, making modular automation attractive where capacity can be redeployed or expanded in stages.

Barrett’s existing UNIT installation uses compact robotic storage and retrieval equipment for piece handling, putaway, inventory movement, fulfilment, and returns. Barrett notes on its own site that supplier performance figures vary by facility, product mix, and volume, an important qualification when assessing the economic case for scaling the system.

Automation offered as a service does not remove the machinery from the equation. Robots, storage structures, workstations, communications equipment, and charging or power infrastructure still occupy warehouse space and require maintenance. The operating model simply changes how some of that capacity is procured and expanded.

Data integrity becomes more important as the system gains authority across the network. An incorrect inventory record affecting one warehouse can create a local problem; inaccurate information used to redirect orders between several facilities can propagate the error more widely.

Integration with warehouse management, order, transport, and customer systems will therefore determine how effectively the automation layer operates. The technology has to receive sufficiently current information to make useful decisions and return reliable status information after physical tasks are completed.

The October announcement does not identify which Barrett locations will form the first 2027 rollout, the number of robots planned, or the proportion of national activity that will initially be coordinated by the platform. Those details will establish whether the first phase is a limited expansion or the beginning of a broader operating architecture.

The significant development is that Barrett is attempting to coordinate automation between facilities rather than simply adding more machinery inside each one. Once deployment begins, the useful measure will be whether that network control improves inventory and equipment utilisation without transferring congestion or complexity into other parts of the logistics operation.