P&G Cuts Shift Planning from 8 Hours to 20 Minutes with Warehouse AI

On November 6, 2026, IIoT World published a video interview with Keith Moore, CEO of AutoScheduler.AI, on the use of artificial intelligence in Procter & Gamble warehouses. The headline figure: shift planning reportedly fell from about 8 hours to 20 minutes. It should be said up front that these numbers are Moore's claims, and the article does not show independent confirmation from P&G.

The problem: a plan that lasts 10 minutes

Across P&G's production campuses, daily workforce planning took about 8 hours per shift. The plan, however, typically fell apart within 10 minutes of the shift starting, because of equipment breakdowns, absences or late trucks. Shuttle movements between buildings must be tightly synchronized to keep lines running and fill rates high: a single delay forced planners to keep recalculating.

The cause: siloed systems

According to Moore, the root cause is traditional warehouse software, built in silos. A site may run a labor management system and, separately, a slotting tool, which rarely talk to each other. Yet labor efficiency depends on where the goods are and when inbound trucks arrive.

The solution: 80% standard, 20% local

AutoScheduler.AI spent six years building a semantic layer, a body of structured knowledge that encodes how a warehouse works optimally. The layer connects WMS and ERP data with algorithms that model equipment capacity and dock door schedules. It automates the universal decisions, which the article puts at around 80% of warehouse operations.

The remaining 20% is specific to each plant: for example, a sorter with particular loading rules, or a cross-dock flow active only on certain days. Instead of an IT development cycle of about 6 months, department supervisors with no programming experience use a no-code AI builder.

According to the interview, results in the field include:

• nearly 500 custom apps built by supervisors, for example to schedule raw-material deliveries to lines and to manage replenishment flows;
• an app for incomplete loads that checks SAP/ERP data on upcoming shipments to the same customer and suggests product substitutions to protect fill rate;
• automation of a manual task that occupied two people, with a claimed saving of $150,000 within two days of release.

Trust comes from explainability

The AI records its reasoning and logic in natural language. An experienced dock supervisor can therefore trace why a recommendation was made, challenge its trade-offs and adjust the parameters. This is key to adoption on the floor, where a black box is unlikely to be accepted.

Key takeaways

Moore's advice is not to get bogged down in endless data cleanup. Map the decisions first: what is being decided, why, and which data informs each choice. Then automate the highest-value areas and abandon fragile spreadsheets.

Decisions first, then data: the map of choices drives what to automate.

The case shows a repeatable pattern: unify siloed systems, automate the standard, and leave local exceptions to the operators.

Open questions

The article does not address several aspects that IT and plant managers should weigh before copying the approach:

• governance of roughly 500 user-built apps: who reviews, versions and retires them;
• the integration effort with WMS, ERP and existing systems;
• the method used to measure times and savings, in particular the $150,000.

Pending independent verification, the numbers should be read as vendor claims. The approach itself, however, is sound: start from decisions, explain recommendations and give tools to the people who know the floor.