Agentic AI Impact: L’Oréal, Cardinal Health & Pet Supplies Plus

MIT recently reported that 95% of enterprises are seeing no measurable P&L impact from their AI investments. Boston Consulting Group came to a similar conclusion — only 5% of companies are generating real value from AI today.

If you're a supply chain leader, you've probably felt exactly this gap. The AI headlines are constant. Internal pressure to "do something with AI" keeps mounting. But when you look at what's actually running in warehouses and distribution networks, most of what you find is pilots that stalled, dashboards that produce insights nobody acts on, and vendor demos that were impressive in the room, but haven't produced the operational lift they promised.

The 5% who are getting value are doing something different. In a recent conversation co-hosted with the National Retail Federation, Aaron Millstone, SVP of Customer Strategy and Applied AI at Manhattan Associates, sat down with four supply chain leaders who are in that 5% — and they were unusually candid about what separates their approach from the 95%.

• Don Bernick, an IT leader at L'Oréal for 33 years, overseeing warehouse and distribution operations across the Americas
• Jagjadish Muka, leader of strategic supply chain IT at Cardinal Health, working with multiple agents already in production
• Jerry Troupe, VP of Logistics at Pet Supplies Plus, running four agents live today with ten more in active development
• Patricio Neri, business expert at L'Oréal for the Latin American zones, leading agent adoption across Uruguay, Argentina, Chile, and Mexico

The panelists don't stay abstract. They talk about the specific workflows where agentic AI is delivering — the replenishment decisions that used to require pulling data from multiple reports, the wave rebalancing that used to fall to a supervisor early in the morning, the exception management that used to cascade before anyone caught it. They also talk plainly about where AI is oversold, why so many enterprise AI investments are failing to move the P&L, and why the deployment cycles they're running are measured in weeks rather than quarters.

The conversation gets specific on questions most vendors would rather avoid:

• What is the 5% actually doing that the 95% is not?
• Where is agentic AI genuinely delivering in warehouse operations today, and where is it still oversold?
• The need for fast deployment cycles, and what operating model between business, IT, and vendor is making that possible?
• What role does human oversight play, and how does that shape adoption on the warehouse floor?