Logistics / Supply ChainNewVerified Real deployment at a named or precisely described company, with a measured result from a primary source.

AI orchestrates a plant-attached warehouse: more loads shipped straight from the plant, fewer shuttle moves (P&G Lima)

Added 2026-10-16 · Last updated 2026-10-16

Role

Warehouse & shipping (plant-attached warehouse) · Typical users: warehouse manager, shipping supervisor, logistics planner, materials handling lead

The problem

A warehouse attached to a plant has to feed raw materials to the lines and ship finished goods at the same time, with limited space, a production schedule that keeps changing and a crew whose size varies from day to day. Supervisors rebuild the plan by hand every day; when the plan is imperfect, finished goods are moved to outside warehouses and later brought back to complete shipments, and shuttle trucks run between buildings all day. At P&G's Lima, Ohio plant, with seven nearby satellite warehouses and more than 250 outbound full-vehicle shipments a day (85% drop-and-hook, 15% live loads), the vendor lists raw-material firefighting to support production, missed shipments, labour coordination, staging and storage limits, travel across the site and siloed data between production, warehouse and logistics systems.

How AI helps

An AI-driven warehouse orchestration platform sits on top of the existing WMS and ERP and reads the production schedule. It plans labour, inventory moves and dock activity in real time against the constraints (dock doors, labour, equipment), sequences raw-material moves to the lines, decides which loads can ship directly from the plant, and re-plans when production changes. Supervisors work from the plan (the case mentions a 'Load Now' button) instead of building it by hand.

Step by step

Map the flows on one page: production output, raw-material staging to each line, storage in the plant and in outside warehouses, docks and shuttle routes.
Measure the baseline for a few weeks: share of loads shipped directly from the plant, bring-backs from outside warehouses, shuttle moves per day, hours spent planning each day, on-time shipment rate.
Pull the data feeds the plan needs into one view: production schedule, WMS inventory by location, open outbound loads with pickup times, dock doors, labour by shift.
Start with a daily draft plan: AI proposes which loads ship directly from plant stock, which need an outside-warehouse pull, the raw-material staging sequence and labour by area; the supervisor approves or changes it.
Re-plan when production changes (a line down, a run extended) and log why each change was made.
Track the baseline measures weekly; if the gain holds, then evaluate an orchestration tool that connects to your WMS and ERP.

Tools

AutoScheduler.AI (warehouse orchestration over WMS and ERP): used at P&G Lima
Other tools to evaluate: the labour management and wave planning modules of your WMS (Manhattan, Blue Yonder, SAP EWM, Körber), dock scheduling software (Opendock, C3 Reservations), yard management systems
Any LLM (ChatGPT, Claude, Copilot) with the prompt below on tomorrow's schedule and open loads, to draft a shift plan and see where the conflicts are

Result

Procter & Gamble, Lima, Ohio manufacturing plant with seven satellite warehouses, using AutoScheduler.AI on top of its WMS and ERP: direct dock shipments from the plant up from 57% to 83.4% (+46.3%); bring-backs from satellite warehouses down from 16% to 8.7% (-45.6%); workforce planning time cut from 8 hours to 20 minutes per day (96%); 30% higher throughput during a product crunch; 99%+ on-time shipment performance; shuttle moves to outside warehouses cut by nearly 50%; more than $4 million in total savings. Source: AutoScheduler's own case study (PDF, undated). The figures are vendor-published; the same PDF also says direct plant-to-customer trucks 'doubled', which does not match the 57% to 83.4% figure, and it gives no time frame for the savings.

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

All figures come from the vendor's own case study; it does not give a time frame for the savings and two of its figures on direct shipments do not match.
A supervisor approves the plan: the AI does not see a forklift down, a quality hold or a driver arriving early unless someone records it.
Do not paste customer names or volumes into a public chatbot; use an approved enterprise tool or remove the identifiers.
Do not let an LLM calculate quantities, capacities or times; check totals in the WMS or a spreadsheet.
Shipping more directly from the plant only works if finished-goods stock is available on time; check that the change does not starve production of staging space.

Try it yourself

You are a warehouse planner at a manufacturing plant with an attached warehouse and [number] outside warehouses.

I will paste: (1) tomorrow's production schedule [line, product, start and end time, quantity]; (2) raw materials needed by each line [material, quantity, needed-by time, current location]; (3) outbound loads for tomorrow [load ID, customer, products and quantities, pickup time, live load or drop trailer]; (4) finished-goods inventory by location [product, quantity, plant or outside warehouse]; (5) resources [dock doors, forklift drivers and loaders per shift, shuttle trucks].

Do this:
1. For each outbound load, say whether it can ship directly from plant stock or from tomorrow's production in time, or needs a pull from an outside warehouse, and why. Copy the quantities from my data; mark anything unclear 'SUPERVISOR CHECK'.
2. Propose a raw-material staging sequence per line so each material is at the line before its needed-by time.
3. Draft a shift plan: which dock door and which crew handle which load or staging task, in time order.
4. List conflicts: two tasks on the same door at the same time, a load that depends on production finishing too late, more tasks than people in a time slot.
5. List the shuttle moves the plan needs and suggest which ones could be avoided by changing the order of work.

Do not add up quantities or calculate capacities yourself: copy the numbers as given and tell me which totals to check in our WMS or spreadsheet. Do not release, book or change any task.
Use only the data I paste in; if something is missing, write 'not found' or list it as a question instead of guessing. Do not send, order, post or change anything in any system: produce drafts for my review and approval.

Want more like this? Join the newsletter

Sources

https://autoscheduler.ai/wp-content/uploads/2025/03/AutoScheduler-PG-Case-Study.pdf