Until AI Can Solve This One Simple Task, I Wouldn’t Worry About It

Warehouse worker crouching down and checking boxes on shelves with scanner. Female worker scanning ... [+] boxes in rack.

Experts have recently warned about the possibility that artificial intelligence could lead to the destruction of 400 million jobs by 2030. These same pundits also predict that AI could lead to increases in global GDP by trillions of dollars in the same time frame. However, AI’s inability to solve the very limited problem of ensuring that inventory is located in the right place in a warehouse suggests that planners don’t have to worry too much about job security. It also suggests that the total value delivered by AI will be more limited than consultants from McKinsey are forecasting.

For fulfillment to be efficient, a warehouse needs the right inventory located in the right slots in a warehouse. Warehouse and supply chain planners rely on a surprising number of technologies to achieve this.

The core system used to manage a warehouse is the warehouse management system. A WMS tells workers what to do. For example, as a picker is moving through the warehouse, their scan gun will read, “go to location XB312 and pick three cases!” The picker arrives at the location, scans the location barcode, which proves they are in the right location, and then scans each case, which proves that the right number of items were selected. Automatic identification solutions—like scan guns or Voice Systems—is what allows a WMS to ensure the right product is picked from the right location. A true WMS is a real-time system that can achieve inventory accuracy of 99.9% or better. Paper-based systems would typically have inventory accuracy in the low 90s. Further, because AutoID is real-time, the WMS always knows how full the pick slots are and which slots need to be replenished.

WMS also achieves its high inventory accuracy through cycle counting. While a worker is doing their tasks, the WMS might tell them to go to slot FD229 and count how many items are in that location. If the inventory number for that slot in the WMS is wrong, the worker updates the system.

When new stock keeping units are weighed and their dimensions are determined, and this information is entered into a WMS, then the WMS understands from that point on how many of those SKUs can fit in a particular slot as well as which racks are strong enough to support heavy SKUs.

Warehouse management systems also employ more advanced functionality for item placement. Slotting logic is used to determine where to put inventory in a warehouse. For unautomated warehouses—those that don’t use robots or miles of conveyors—dynamic slotting is the way to make sure inventory is in the right place. Dynamic slotting is based on the idea that the less total walking there is, as workers go from one pick slot to the next, the more productive that warehouse will be. Overall, this leads to goods getting fulfilled more quickly and economically. In dynamic slotting, goods for which there is high demand are located nearer to the shipping docks to minimize travel.

Sometimes the right place for inventory isn’t in the storage slots in the warehouse. It is better to receive inventory on a loading dock, take the inventory needed for a hot shot shipment and move that inventory through the DC to a shipping dock where it is loaded on a truck. This is known as “flow through.” Our ability to do this has gotten better because real-time transportation visibility solutions give the warehouse much better certainty surrounding when a truck will arrive at the DC.

But just because needed inventory arrives in time to be placed on an outbound truck does not mean that this will happen. The inbound truck may need to go to the front of the line; A yard management system needs to be integrated to the WMS to make this happen. The truck might have cubed or weighed out and the WMS does not know this. Flowthrough needs planning that spans across both warehouse and transportation management systems.

While a WMS is the core solution for managing the warehouse, other applications also play a large role in making sure inventory is in the right place. First, inventory can’t be in the right place in a warehouse if the warehouse is not in the right place. Network design is used for this. Supply chain design solutions look at long-term demand projections and then calculate how many warehouses are needed, and where those warehouses need to be located, to meet targeted service levels at the lowest possible cost.

Inventory optimization is another necessary precursor solution. An IO solution uses sophisticated math to determine how much inventory needs to be stored at different locations across the network to make certain service levels will be attained with the lowest possible amount of inventory.

In an omnichannel world, companies have different flow paths to get products to consumers—buy in store, ship from a DC, ship from store, etc. A distributed order management system determines which warehouse or store should be the originating location for an order.

Just determining the right location for inventory in a warehouse requires multiple systems and technologies. This article showed the role eight different technologies play. Without AutoID, it doesn't matter how clever the AI is, it cannot effectively solve the location problem. In short, you need the right building blocks in place to get value out of AI.

But the job of planners is not just to have the right inventory in a warehouse. It is to have the right product, in the right place, at the right time, across an extended supply chain. There are at least another six applications that would have a role to play in achieving this goal.

I don’t worry about AI eliminating tens of thousands of planning jobs while simultaneously driving unprecedented levels of productivity. There are just too many point solutions. As things stand, AI can improve a particular supply chain function, but that doesn’t holistically optimize the system.

Let me explain this. Companies can invest to make sure their factories are efficient. They can do the same for their distribution and transportation functions. If each function is as efficient as possible, doesn’t this mean the system would be working at peak efficiency?

No! It doesn’t. It may be that the factories are achieving good quality at a low unit cost, but they are doing that by storing too much raw material inventory in the inbound warehouses and by having long product runs that produce too much unneeded inventory. The factory achieves its key metrics, but the metrics that really matter, a high service level at the lowest possible cost, are not met.

Until a more holistic approach to supply chain software is achieved, an approach that drastically collapses the number of models used in planning, neither the potential nor the threats associated with AI will come fully into play. It is not that we can’t do this, but this will require a high degree of investment, innovation, experimentation and ultimately many, many years of effort.