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

On 6 November 2026, IIoT World published a video interview in its CxO series with Keith Moore, CEO of AutoScheduler.AI, about the use of AI in Procter & Gamble warehouses. The headline figure: daily workforce planning, which used to take about 8 hours per shift, now takes 20 minutes.

A necessary caveat: the source is a vendor CEO's own account. The article was summarized with AI tools and reviewed by our editorial team, but it offers no independent verification of the numbers. They should be read as indications, not as validated benchmarks.

The Problem: A Plan That Lasts Ten Minutes

On P&G's production campuses, the work plan, built over roughly eight hours, usually fell apart within ten minutes of the shift starting. The causes will be familiar to anyone working in logistics: equipment failures, sick leave, late trucks. Shuttle transfers between buildings have to be tightly synchronized, so a single delay forced constant replanning.

According to Moore, the root cause is siloed legacy software: a labor management system and a separate slotting tool rarely talk to each other, even though labor efficiency depends on where the goods are and when the trucks arrive.

The Solution: A Semantic Layer and a No-Code Builder

AutoScheduler.AI spent six years building a semantic layer: a unified body of structured knowledge that encodes how a warehouse operates optimally. It connects WMS and ERP data with algorithms that understand equipment capacities and dock schedules. This automates the universal decisions, about 80% of operations.

The remaining 20% is site-specific: a sorter with particular loading rules, a cross-dock flow active only on certain days. Instead of an IT development cycle of about six months, the platform gives supervisors a no-code AI builder. With no programming experience, they have created nearly 500 custom apps, for example to schedule raw material deliveries to specific lines or to handle particular replenishment flows.

One app mentioned checks, when a load is short of product, SAP and ERP data on upcoming shipments to the same customer and proposes substitutions to protect the fill rate. It replaced a manual task that tied up two people; the supervisor who built it reportedly saved his site $150,000 within two days of release.

Explainability as an Adoption Lever

The AI acts as a transparent decision-making agent: it records its mathematical reasoning in plain language. A dock supervisor with twenty years of experience can reconstruct why a recommendation was made, challenge the trade-offs and adjust the parameters.

A system that operators can question and correct is more likely to be used.

Key Takeaways

Three lessons for those working in automation and industrial software:

• Two-tier architecture: standardized core logic, with edge cases configurable locally by the people who know the process.
• Explainability: readable rationales and adjustable parameters reduce operators' distrust.
• Map decisions first: Moore's advice is not to get lost in endless data cleanup, but to identify which decisions are made, why and with what data, automate the most relevant ones and abandon fragile spreadsheets.

A personal note: the article mentions neither discrete-event simulation nor digital twins. Still, a plan that breaks within ten minutes suggests that testing delay and absence scenarios before the shift would be a natural complement to the optimization. This is our own observation, not something from the source.