Warehouse AI Agents Now Make Inventory Calls
A survey of nearly 2,000 supply chain professionals finds agentic AI moving into daily operations, though most reported gains remain unquantified.
Updated on Nov. 17, 2026 in Artificial Intelligence

Live Poll
Do you think now is a good time for AI to make more of your daily decisions?
MIT's Center for Transportation and Logistics and warehouse-systems firm Mecalux released a global report on generative and agentic AI in warehouse operations, based on a survey of nearly 2,000 supply chain and warehousing professionals across 21 countries. The findings describe deployment in research-stage survey terms as reported by respondents, with 61% of organisations reporting GenAI at large-scale deployment or full integration and 56% reporting agentic AI at that stage.
Why it matters
The report shows agentic AI moving from pilot into operational control, with autonomous decisions already common in inventory and data work. Because about three in four AI applications have been at their current stage for less than a year, the report points to a fast-moving field whose governance and measured returns are still maturing.
The figures come from a survey of nearly 2,000 professionals across 21 countries, in which 87% of GenAI and 81% of agentic use cases showed at least moderate improvement in target metrics. Only 39% of GenAI and 36% of agentic use cases had quantified, documented significant gains, so most reported improvements remain unmeasured.
The players
Massachusetts Institute of Technology
MIT is a leading research university whose Center for Transportation and Logistics produces applied research on supply chains and logistics systems.
Mecalux
Mecalux is a warehouse storage and automation systems company that co-released the global report with MIT.
The details
Agentic AI refers to software agents that take actions on their own rather than only generating content. According to the survey, agents make decisions without human approval on inventory movements in 51% of agentic use cases and on data corrections in 42%. The most common oversight model is human-on-the-loop, in which the AI operates within defined limits while staff monitor and can intervene. Businesses with comprehensive, actively enforced AI governance are about twice as likely to report significant, documented improvements.
Timeline
2028: organisations expect AI to autonomously make or execute more than a quarter of warehouse decisions.
Over the next two to three years: 90% of organisations expect to increase AI investment.
The report recommends reviewing AI application results within the first 12 to 24 months.
The Tech Race
The report sits within a broader push by logistics researchers and warehouse-automation vendors to define how agentic AI is governed on the operations floor. It is a survey of practitioners rather than a controlled benchmark, so it documents adoption and perceived results rather than a head-to-head performance ranking against competing systems.
Warehouse operators and their staff are most directly affected, since agents already make inventory movement decisions without human approval in about half of reported agentic use cases. Customer-facing commitments are permitted in 22% of agentic use cases and financial decisions in 12%, so oversight design determines how much of that autonomy reaches customers and budgets.
The takeaway
The report suggests the decisive variable in warehouse AI is governance rather than model capability, since businesses with enforced oversight are about twice as likely to document significant gains. Readers should watch whether the 2028 projection of AI making over a quarter of warehouse decisions is revisited in the report's follow-up measurements over the next 12 to 24 months.
What happens next
Organisations expect agentic AI to make over a quarter of warehouse decisions autonomously by 2028, a milestone the report's respondents will be measured against in the years ahead.
Further reading
For broader context on how these systems are being built and governed, see the Artificial Intelligence section.
More information
Download the full MIT-Mecalux AI warehousing report from Mecalux.
Live Poll
Do you think now is a good time for AI to make more of your daily decisions?





