Artificial intelligence is gaining a foothold across warehouse operations, but the more revealing divide is not between adopters and non-adopters. It is between systems that surface information and systems trusted to act on it.
New data from Logistics Reply shows that warehouses are using AI across reporting, replenishment and operational response, while keeping people firmly inside the decision loop. That creates a different challenge for the next phase of deployment. Greater automation will depend on whether organizations can establish the data, controls and system connections required to give AI more operational authority.
AI Is Strongest Where the Decision Risk Is Lower
Nearly half of respondents, 49%, said AI is running in at least one warehouse process, being piloted in multiple areas or embedded across much of their operations. A further 28% are evaluating where AI could add value, while 23% have not started engaging with the technology.
Current use cases show where organizations are most comfortable deploying it. Reporting and analytics was the most common application at 32%, followed by slotting or replenishment at 27%. Order fulfillment and yard or dock management were also among the areas where AI is being applied.
Yet adoption does not necessarily translate into decision authority.
When respondents were asked what AI currently does inside their warehouse, 31% said it produces reports, dashboards or operational insights. Only 3% said AI makes operational decisions independently.
The same pattern appears when warehouses encounter disruption. Roughly 48% said AI identifies an inventory shortage, labor gap, equipment failure or similar problem while employees manually coordinate the response. A nearly identical share said AI recommends the next action but leaves the final decision to people.
Only 4% said AI automatically coordinates a response across systems.
That distinction matters because operational visibility creates value without necessarily transferring control. Respondents identified better operational visibility as AI’s largest business contribution at 33%, followed by reducing manual or repetitive work and identifying problems earlier.
The Next Constraint Is Operational Authority
Trust remains a significant barrier to deeper automation. Some 83% of respondents said they usually review AI recommendations before acting on them. Only 5% trust AI sufficiently to act on recommendations without review.
Human intervention is also built into existing governance. When AI makes a recommendation or decision, 75% said employees can override it, either subject to approval or depending on the process involved.
The technology architecture creates another constraint. While 58% said AI coordinates activity across several connected operational systems, 35% said it primarily operates within one application. Only 1% reported AI orchestrating workflows across multiple systems such as warehouse management, enterprise resource planning, transportation, labor and automation platforms.
That integration gap becomes increasingly important as organizations pursue more autonomous operations. An AI system may identify that inventory is unavailable or labor capacity is insufficient, but executing a response can require coordinated changes across replenishment, task allocation, transportation and automation systems.
The survey points directly toward that ambition. Autonomous decision-making was identified by 30% of respondents as AI’s greatest untapped opportunity, ahead of inventory optimization and labor planning. Asked what capability they wanted next, 23% selected recommendations for the best operational action, while 22% selected orchestration of decisions across multiple systems.
Getting there will require more than increasingly capable models. Budget constraints were identified by 25% as the biggest obstacle to extracting greater value from AI, followed by data quality at 22%. Difficulty demonstrating return on investment and shortages of internal expertise were also cited.
These barriers are interconnected. Poor data limits confidence in recommendations, limited integration restricts execution and persistent human checking can reduce the labor and speed benefits that greater autonomy is expected to deliver.
Autonomy Changes the Cost of a Bad Decision
As warehouses give AI authority to trigger replenishment, redirect work or coordinate connected systems, performance measurement will need to extend beyond labor savings and faster decisions. Operators will need to track how often automated actions are reversed, how exceptions propagate into downstream processes and where human intervention still prevents costly errors. Those measures can help determine which decisions are ready for greater autonomy and which still justify the cost of human review.
