Last Mile Delivery and the WMS Gap: Why Warehouse Automation Stalls Without an Orchestration Layer

The bottleneck sits past the loading bay

Every intralogistics roadmap conversation in Europe runs on the same axis: high-speed sorters, autonomous mobile robots, goods-to-person systems, and the next generation of warehouse management software. The investment case for all of it rests on throughput.

The uncomfortable part is that European consumers are not experiencing the throughput. Eurostat found 35.4 percent of EU online shoppers reported a problem in 2025, with the most common complaint being slower-than-expected delivery at 19.9 percent. Warehouse automation has advanced substantially over the same period. The customer-facing outcome has not moved with it.

The reason is structural rather than technical. You can compress pick-and-pack from two hours to fifteen minutes, and if the resulting orders sit on a staging dock waiting for a carrier without capacity, the fifteen minutes have been converted into dock dwell rather than into delivery speed. The cost concentration confirms where the outcome is decided: Capgemini Research Institute puts last-mile delivery at 41 to 53 percent of total logistics and shipping cost.

The gap between the two is a handover, and handovers are expensive in a way that appears on no budget line. McKinsey estimates that inefficient logistics handovers account for 13 to 19 percent of logistics costs, as much as 95 billion dollars annually in the US alone. The warehouse-to-carrier handover is one of the densest in the chain, and it is the one automation programmes most often leave untouched.

Also Read: Warehouse-to-Doorstep Latency: Where European Fulfilment Automation Actually Loses the Speed Race in 2026

Islanded efficiency: what a WMS cannot see

A WMS is very good at what it was built for. It batches orders to minimise travel paths, manages inventory locations, and directs human and automated workflows against available labour.

What it cannot do is see past the dock door. When order release is driven purely by internal capacity, three failure modes follow.

Staging dock congestion. Orders are picked ahead of the carrier’s arrival and accumulate on the dock floor. Beyond the operational disorder, product spends longer in the least controlled part of the building, which matters more with every temperature-sensitive or high-value line.

Missed carrier cut-offs. High-priority parcels are picked in the same waves as standard ones because the WMS has no reason to distinguish them. The order that needed to make an 16:00 departure is ready at 16:20, and a day is lost to a twenty-minute miss.

Capacity mismatches. The site picks volume that the receiving parcel hub or micro-fulfilment node cannot inject into that day’s delivery routes. The warehouse records a productive shift while the network records a backlog.

None of these are WMS defects. They are the predictable result of optimising one system against a constraint set that excludes the next system.

Why WMS and a traditional TMS together still leave a gap

The common assumption is that adding a TMS resolves this. It does not, because the two systems optimise different objectives and the interface between them is passive: orders pass sequentially rather than being released in response to what is happening downstream.

The missing function is an active controller between them. Without it, both systems perform well against their own metrics while the outcome that matters, on-time in-full delivery, is set by whichever of them is wrong at the dock that morning.

This is also where most European operations sit today on execution capability generally. Gartner found that while 95 percent of supply chains must react quickly to change, only 7 percent can execute decisions in real time.

Also Read: TMS-WMS-ERP Integration Architecture for Enterprises in 2026

What an orchestration layer actually does

An orchestration layer sits between the inventory engine, the WMS, and the last mile carrier network, and it continuously evaluates external conditions in order to govern internal workflow. Three functions carry most of the value.

Dynamic wave and order release. Rather than releasing orders in fixed time batches, the orchestrator evaluates live carrier capacity, traffic conditions, micro-hub backlog, and courier cut-off times, then determines when the WMS should trigger the pick. The pick happens when the network is ready to receive it, not when labour happens to be free.

SLA-driven re-prioritisation. When a regional courier hits route delays, the affected parcels are re-prioritised in the warehouse queue automatically so they make an earlier departure window. This is the function that converts a downstream problem into an upstream adjustment rather than into a missed promise.

Multi-node realignment. For operations running several distribution centres, urban micro-fulfilment sites, or store fulfilment nodes, the orchestrator can reassign an order to a different node mid-process when the original node’s last mile capacity is congested. The decision moves to where the delivery can actually happen, before any pick cost has been sunk.

There is a European efficiency dimension to the third function specifically. Eurostat reports that 21.6 percent of distances travelled by road freight vehicles in the EU were performed by empty vehicles in 2024, rising to nearly 26 percent for national transport. Node assignment made without reference to network position is one contributor to that number.

Also Read: Multi-Carrier Orchestration: How AI Intelligent Order Allocation Reduces Enterprise Shipping Costs in 2026

How this works on Locus

Locus, the world’s first Decision-Intelligent, Agentic TMS, operates as this layer alongside existing systems. The WMS and ERP remain systems of record; Locus becomes the system of execution. Built by Mara Labs Inc. and acquired by Ingka Group, parent of IKEA, in 2025, it runs 250+ real-world constraints per computation across 360+ enterprise customers and 30+ countries, and is ranked #1 in Route Planning on G2’s 2026 Best Software Awards.

Mechanically, DiSCO runs eight named agents on a continuous Sense, Decide, Execute, Learn cycle. The Hub agent manages facility readiness and the dispatch handover, the Capacity agent forecasts demand and matches available capacity, the Carrier agent allocates across the network, and the Dispatch agent sequences and re-sequences as conditions change. Six governance mechanisms bound autonomous action, including autonomy levels and human-in-the-loop override, so an operation can start with release recommendations before moving to autonomous wave triggering.

Also Read: Stop Routing Bad Promises: Why Last-Mile Efficiency Actually Starts at the E-Commerce Checkout

Two deployments show the warehouse-to-dispatch seam closing. A global FMCG leader had the exact pattern described above in reverse: picking stalled at the warehouse waiting for manual plans, with scheduling cycles running long against tight SLAs. Autonomous planning collapsed a three-hour manual cycle into a five-minute run, clearing the picking bottleneck, with the Hub agent readying shipments through no-touch planning and 12,000+ trips eliminated each month. A leading ASEAN apparel retailer went further upstream, creating labels and shipments at the point of packing with carrier label generation in under 500 milliseconds, so warehouse throughput no longer waited on a carrier portal or a manual step, alongside 99 percent-plus delivery SLA.

Also Read: Beyond the Highway: Why Real-Time Visibility is Shifting Focus to Yard Management and Dock Orchestration

What to ask before the next automation investment

Automation inside the four walls is half the problem. As labour costs rise and narrow-window delivery becomes the European baseline rather than a premium, internal velocity has to be synchronised with external capability.

Four questions worth putting to any intralogistics vendor, and to your own operations team, before the next capital request.

1. What triggers order release in your system, and can that trigger be an external condition rather than internal labour availability?
2. When a carrier cut-off is at risk, what changes in the pick queue, and does it change automatically?
3. If a fulfilment node’s last mile capacity is saturated, can an order be reassigned before it is picked, or only after it is packed?
4. What does the site do differently on a day when the carrier arrives 90 minutes late?

The last one is the most revealing. Most highly automated sites have no answer, because nothing in the building knows the carrier is late until the driver arrives.