)
What Warehouse Leaders Would Cut If They Could Do It Again
The second section of most good industry panels is usually when things get really interesting. The first part lets everyone settle in and say the right things. By the second, people start saying what they actually think.
In a recent panel discussion hosted by Logistics Reply, three veterans with close to a century of combined warehouse experience were asked to identify the tools, approaches, and investments they would kill. What followed was a candid set of stories about technology that failed in practice, decisions that should have been stopped earlier, and a few industry-wide habits that may be doing more harm than good.
This is the second post in a three-part series. If you missed part one, it covers where the panel sees genuine, durable value worth protecting. This post covers what they would cut.
The Automation That Ran Without an Off Switch
The warehouse systems director opened with a story that made the point better than any statistic could.
Her company had implemented an automated storage and retrieval system for full pallet picking at one of their facilities. The system would pick pallets and drop them to a staging lane for outbound shipment. Under normal conditions, it worked well. Then one day a driver arrived for a pickup, and the system got stuck. Two pallets remained to be staged, but instead of completing the job, the LGVs carrying them kept circling the warehouse in a loop, unable to resolve whatever exception had stopped them from dropping to the lane. There was no interception point in the process, no mechanism for a human to step in and take over. The driver sat there for three and a half hours watching machines go around. The problem was not automation. It was that nobody had designed for what happens when the algorithm hits a situation it cannot resolve on its own.
"AI is wonderful when it is repeatable, when you can write an algorithm for it and it doesn't deviate. But there always has to be a point where a human can step in."
Her framing was precise: automation earns its value in high-volume, repeatable processes. The moment you push it into exceptions, into the edge cases and the situations that do not fit the pattern, you need human judgment available. Interception points are not optional, but a necessary part of the design.
The Voice Picking System Without a Backup
The CPG veteran described a voice picking implementation that looked compelling in the demo and created real operational pain in practice. The facilities were smaller, under 100,000 square feet, serving independent distributors. The picking platform was well-regarded. But the system ran on software that required periodic security credential renewals, and when one expired, the whole thing went down. No one had purchased handheld barcode scanners as a backup. Workers had nothing to fall back on. The operation was down for 12 hours and orders did not ship.
The lesson he drew was not that voice picking is a bad technology. It is that a technology selection cannot be evaluated in isolation from the question of what happens when it fails. A modest investment in backup devices would have kept the operation moving. The demo never showed that scenario.
Early AI Implementations Under Pressure
The consultant was also direct about what he is seeing with early AI deployments across major organizations. Companies with large, sophisticated IT departments are adopting AI systems that promise smarter decision support, better operational visibility, and faster, more intelligent responses to exceptions. Many of those systems are not lasting.
The early adopters get compelling demos. The AI system is implemented. And then the gap between what the system promised and what it actually delivers in a live operational environment becomes clear. The kinks are real; the fixes take longer than the business can wait, and the systems get pulled out. Organizations that moved fast to adopt are now moving fast to remove.
This is not a permanent verdict on AI in the warehouse. The panel returned to AI in depth in the third section of the discussion. But the message here was specific: be cautious about first-mover adoption of AI tools that have not been stress-tested in environments like yours. The demo and the live operation are two different worlds. The consultant named a pattern he sees across companies of every size, including some of the largest and most sophisticated operations in the world.
A company commits $200,000 to a technology implementation. It becomes clear midway through that the technology is not going to deliver what was promised. The cost to complete ballons to $400,000. And they still run it all the way through anyway.
The Sunk Cost Problem
The consultant named a pattern he sees across companies of every size, including some of the largest and most sophisticated operations in the world.
A company budgets $200,000 for a technology implementation. Midway through, it becomes clear the technology is not going to deliver what was promised. At that point they have two options: stop and absorb the loss or finish the project at a total cost of $400,000. Most choose to finish. They end up spending twice their original budget to complete something they already know will not work.
The psychological logic is familiar. Stopping feels like failure. But the money already spent is gone either way. Finishing does not recover it. It just doubles the loss.
His prescription was blunt: define the exit criteria before the project starts, and get agreement at every level from the capital committee down. Organizations that are disciplined about this protect themselves from one of the most common and expensive failure modes in supply chain technology investment.
"I must be able to quit when it's time to exit. That kill button has to be well-defined, and everyone has to agree on it from the start."
Six Sigma and Siloed Problem-Solving
One of the more unexpected kill recommendations came in a conversation about methodology rather than technology. The consultant described how Six Sigma, for years the default framework for attacking operational problems, is increasingly ill-suited to modern supply chain environments.
The traditional Six Sigma approach works by isolating a problem, focusing resources on solving it, and moving to the next one. That model made sense when systems were more discrete. But today, supply chains are deeply interconnected. A warehouse does not just hold its own stuff. It sits inside a flow that starts at demand planning and ends at the customer. Shooting a siloed problem in the head, without understanding how the fix propagates upstream and downstream, creates new problems as fast as it solves old ones.
The consultant pointed to systems thinking as the emerging framework; a set of ideas built around understanding problems holistically across the full flow rather than in isolation. The shift is gaining traction, and the panel noted that academic voices in the room were likely to recognize the reference.
Over-Complicated Configurations
The warehouse systems director brought the kill conversation back to a practical place with a point about configuration discipline. When evaluating software vendors, her team regularly encounters implementations that have accumulated years of customizations. Some of those customizations addressed real business needs. Many of them addressed needs that no longer exist, or that could have been handled with standard configuration.
The result is systems that are very difficult to upgrade, because the vendor's upgrade path does not account for the custom layer. As software companies accelerate their migration to cloud-based architectures, that customization debt is becoming an acute problem. Vendors are telling customers to move to a new version or pay a significant fee for continued support on the old one. Companies sitting on heavily customized on-prem systems are finding out that the cost of those customizations was deferred, not avoided.
Her recommendation was to be precise about what you actually need. Not what the vendor can configure, but what your operation requires to do its job. The gap between those two answers is usually the source of the problem.
Coming Up Next
The third and final post in this series covers what the panel would add. Digital twins, integrated planning, AI that actually delivers, forklift driver assist, and their predictions for where warehouse operations will be by 2028. Some of them are closer than you think. Some of them are already here.
How Logistics Reply Helps
Many warehouse challenges are not caused by a lack of technology. They are caused by excessive customization, disconnected systems, and processes that have become difficult to maintain.
LEA Reply was designed with a modular, microservices-based architecture that allows organizations to add capabilities as needed while minimizing customization and upgrade risk. The result is greater flexibility, faster innovation, and a smoother path to modernization.
Explore how a composable supply chain execution platform can help simplify operations and reduce long-term complexity.
Contact Us and one of our supply chain experts will reach out to learn more about your business and warehouse execution challenges.