AI routing and warehouse intelligence that cut 3PL processing time 57% across 4,200 daily orders

Industry

Logistics

Market

India

Client

Third-party logistics provider in India

Sponsor

CEO

What we built

Order routing and warehouse intelligence

The situation

The client is a third-party logistics provider in India running three warehouses and handling 4,200 orders a day. Those orders were being routed and planned manually: people deciding which warehouse fulfils which order, how orders are grouped for picking, and in what sequence the floor works through them.

Manual planning works at low volume. At 4,200 orders a day it becomes the bottleneck. Decisions are made on habit and incomplete information, orders are sent to warehouses where stock is thin, pickers walk further than they need to, and the day's capacity is set by how fast the planners can think rather than how fast the floor can move. Growing meant hiring more people and finding more space.

What we built

A decision layer on top of the client's existing warehouse operations that makes the planning calls automatically and keeps making them as the day changes.

Order routing. Each incoming order is assigned to the warehouse best placed to fulfil it, weighing stock position, distance to the delivery destination, the warehouse's current workload and the order's service level. Split shipments are avoided where possible and chosen deliberately where they are cheaper or faster.

Batching and waves. Orders are grouped into picking batches and waves that share locations and carriers, so each trip through the aisles fulfils more orders and each outbound dispatch is fuller.

Pick-path sequencing. Within a batch, picks are sequenced to minimise walking distance through the actual layout of each warehouse.

Replanning through the day. When stock runs short, a carrier cutoff approaches or a warehouse falls behind, the plan adjusts rather than waiting for the next manual review.

Visibility for operations. Supervisors and the leadership team see orders, backlog and throughput by warehouse in one place, with the reasoning behind each routing decision available when they want to check it.

The system works through the client's existing warehouse management system rather than replacing it, so floor teams kept the tools they knew.

How it went live

We learned the operation before automating it: historical orders, stock movements and layouts for each warehouse. The routing and batching logic ran first as recommendations that planners could accept or override, which surfaced the operational rules that were never written down. Once recommendations were being accepted consistently, the system moved to making the calls directly, one warehouse at a time.

Results

Processing time fell by 57%. With the same staff and the same floor space, the operation now achieves nearly double the throughput. In the CEO's words, they did not think the numbers were real until they saw them.

What made the difference

Improving decisions, not replacing systems. Working through the existing WMS kept the floor stable while the planning got smarter.

Recommendations before automation. Letting planners override the system early is how the unwritten rules got captured.

Optimising the whole day. Continuous replanning, rather than a single morning plan, is where much of the throughput came from.