Some warehouses in China already operate with four to five people per shop floor, the so-called ‘dark warehouse’. Amazon has deployed more than 750,000 mobile robots across its fulfillment network. Ocado’s Erith facility in the UK uses 3,500 robots to process over 200,000 orders per week. The technology exists. The question is less whether full automation is possible and more whether it makes sense for your operation, and when.
Who are we? Speed Commerce is an end-to-end provider of outsourced customer experience solutions for eCommerce retailers (including for Shopify and BigCommerce and more) as well as manufacturers, for close to 20 years. We grow our clients’ businesses by providing winning customer experience strategies such as 24/7/365 eCommerce customer service, order fulfillment, and warehousing – get a free quote from a fulfillment expert. Use, share & embed our new fulfillment logistics calculators optimized for 2026
What does ‘fully automated’ mean?
Before we can answer when warehouses will be fully automated, we need to agree on what that means. The industry uses the term in three distinct ways:
High automation means robots handle travel, storage, sortation, some picking, and some packing. People still manage exceptions, replenishment, maintenance, returns, damage assessment, and oddball SKUs. This is what most modern warehouses aim for.
‘Lights-out’ or near-full automation means the normal inbound-to-outbound flow runs with minimal human involvement. People still show up for maintenance, safety checks, jams, replenishment, cycle counting, and edge cases where the system encounters something it can’t handle. Think of it as 90% automated during standard operations.
Full automation across the industry would mean the average third-party logistics warehouse handling mixed SKUs, kitting, returns, promotional builds, and shifting client requirements runs with essentially no human labor. This is what people picture when they ask the question, and it’s the least realistic scenario.
The gap between these definitions matters. A warehouse can be ‘highly automated’ and still employ hundreds of people. A ‘lights-out’ facility still needs humans on standby. True industry-wide full automation is more concept than reality.
Where does warehouse automation stand right now?
The warehouse automation market reached $34.17 billion in 2026, up from $19.23 billion in 2023. That’s significant growth, but it tells only part of the story.
Interact Analysis projects that 26% of warehouse sites will have some form of automation by 2027. That’s up from 14% in 2017 and 18% at the end of 2021. The trend points upward, but slowly. It also means that by 2027, three out of four warehouses will still operate without automation.
The numbers get more interesting when you look at what ‘some automation’ means. A 2026 survey found that labeling is automated in 24% of warehouses, reporting in 22%, and inventory control in 20%. Picking and packing, the most labor-intensive activities, sit at just 15-18% automation. These are the tasks that drive the vision of lights-out warehouses, and they’re proving the hardest to automate at scale.
The mobile robot market grew 27% in 2023 to reach $4.5 billion. Forecasts initially projected the market would exceed $14 billion by 2027. Those projections have since been revised to $15.6 billion by 2030, with the five-year compound annual growth rate trimmed to around 21%. The revision signals what many operators already know: implementation takes longer than expected, costs more than budgeted, and delivers returns more gradually than hoped.
Is software moving faster than hardware?
The narrative gets a little convoluted here. While physical automation penetration sits at 26% by 2027, research from MIT’s Center for Transportation and Logistics and Mecalux found that nine out of 10 organizations deploy artificial intelligence or machine learning in warehousing operations. AI is integrated in roughly 60% of warehouses, with reported return on investment timelines as fast as two to three years.
Warehouses are digitizing faster than they’re robotizing. Software that optimizes routes, predicts demand, manages inventory, and coordinates labor can be deployed in weeks or months. Physical automation systems take months or years to implement and require significant capital investment.
A warehouse management system with AI-powered optimization can deliver measurable improvements without requiring a single robot. That’s why 90% adoption for AI and 26% adoption for automation aren’t contradictory statistics. They represent different strategies for different problems.
What can technology automate versus what actually gets automated?
McKinsey research makes the distinction clear: currently demonstrated technology could automate approximately 45% of the activities people are paid to perform across the economy. In predictable physical environments, that technical feasibility rises to 78%.
Warehouses qualify as predictable physical environments. So why aren’t they 78% automated?
Because technical feasibility and practical deployment are different things. Cost, labor economics, quality considerations, and organizational acceptance all shape what actually gets automated. A solution might work perfectly in a controlled test but fail when faced with the chaos of real-world operations.
Third-party logistics providers face additional complications. Equipment life cycles span seven to 10 years, but client contracts might run two to three years. Investing in automation for a client who might leave creates financial risk. Technology keeps evolving, making it hard to know which systems will become industry standards and which will become expensive orphans.
Omnichannel volatility compounds the problem. A warehouse designed for bulk shipments to retail stores operates differently than one handling individual e-commerce orders. Many facilities now do both, sometimes for the same client. Automation designed for one model struggles with the other.
What do the most automated warehouses look like?
Ocado runs some of the most automated grocery fulfillment operations in the world. The Andover facility uses over 1,000 robots to process 3.5 million items and 65,000 orders per week. The Erith facility spans 560,000 square feet, employs 3,500 robots, and processes more than 200,000 orders weekly. The system can pick and pack a 50-item order in five minutes. Robots travel at four meters per second, passing within five millimeters of each other.
Ocado has been operating highly automated warehouses for more than 20 years. The company stores 78% more products than a typical supermarket in the same space. This is what mature, purpose-built automation looks like when you control the entire operation, from warehouse design to product packaging to order profiles.
Amazon is a different model. The company acquired Kiva Systems in 2012 for $775 million and has since deployed those robots across its network. Financial reporting suggests Amazon expected approximately $100 billion in capital expenditures for 2025, with up to $25 billion directed toward retail network modernization using robotics. At one next-generation site, the company reported a 25% fulfillment cost reduction.
The most aggressive operator in the industry, with more than 750,000 mobile robots deployed and billions invested in automation, is talking about 25% cost reduction, not elimination of human labor.
Even Walmart, after piloting its Alphabot system in a limited rollout, ended its contract with Bossa Nova Robotics in 2020. The system worked for certain tasks but still required humans for fresh product picking and other variable activities.
Why does picking and packing still resist automation?
At 15-18% automation rates, picking and packing represent the bottleneck preventing true lights-out operations. The reasons are both technical and economic.
SKU chaos and packaging variance create problems that robots struggle to solve. Mixed-case picking, different-sized items, deformable packaging, damaged cartons, unclear labels, and vendor inconsistencies all complicate automation. A human picker can adapt on the fly. A robot needs specific parameters.
Returns amplify the challenge. Grading returned items, inspecting for damage, making refurbishment decisions, and re-bagging products all require judgment calls. The process involves too much variation for current automation to handle reliably.
Kitting and promotional builds add another layer of complexity. These tasks involve high variability and frequent changeovers. The economics rarely justify automating a process that changes every few weeks.
Exception handling lies firmly in human territory. When robots jam, items get mis-picked, cycle counts don’t match, or inventory exists in the system but not on the shelf, people have to intervene. As automation increases, the exceptions don’t disappear. They simply become more visible and more disruptive when they occur.
What’s driving automation despite these challenges?
Labor economics and safety concerns push companies toward automation even when the technology isn’t perfect.
The Bureau of Labor Statistics shows that warehousing and storage (NAICS 493) had a total recordable case rate of 4.8 per 100 workers in 2024. That’s more than double the 2.3 rate for private industry overall. Injury rates that high create both moral and financial pressure to remove people from the most physically demanding tasks.
E-commerce growth intensifies the pressure. The e-commerce market reached $6.3 trillion in 2024. Processing individual orders for consumers requires different workflows than processing bulk orders for retail stores. Labor-intensive picking and packing don’t scale the same way automated storage and retrieval does.
The math can work even with partial automation. Autonomous mobile robots deliver payback in under 24 months with return on investment above 250% in many applications. That’s compelling enough to justify deployment even if the robots only handle specific tasks rather than entire workflows.
When should we expect widespread automation?
Gartner predicts that by 2027, over 75% of companies will have adopted some form of cyber-physical automation. That’s not full automation. That’s ‘some automation,’ which could mean anything from a few mobile robots to a comprehensive automated storage and retrieval system.
McKinsey suggests that by 2030, most warehouse operations could automate simple and repetitive tasks as AI takes over those functions. Again, this describes task-level automation, not facility-wide lights-out operations.
A McKinsey retail study found that over the next five years, only 20% of retailers expect to automate five or more operations to a very high degree. Some of those will build fully automated warehouses. Another 44% plan to automate one to four operations significantly. The remaining 38% have not specified any operations as highly automated priorities.
These numbers describe a measured, selective approach to automation. Companies are choosing specific processes to automate rather than pursuing blanket automation strategies.
What timeline makes sense for different scenarios?
Looking at the data, the research, and the real-world implementations, three scenarios emerge with different timelines.
From 2026 to 2030, expect rapid spread of AI optimization tools and selective robotics deployment. Autonomous mobile robots, sortation systems, and automated storage and retrieval systems will become common in higher-throughput facilities. Most warehouses will have some automation, but few will approach lights-out status.
In the early to mid-2030s, more facilities will achieve near lights-out operation in specific flows, particularly in highly standardized operations like parcel sorting, certain grocery applications, and some manufacturing distribution centers. These facilities will still require staff for exceptions and maintenance, but normal operations will run with minimal human intervention.
From the late 2030s into the 2040s, a growing share of new-build, high-volume fulfillment centers may operate with very low labor hours per unit processed. However, full automation as the norm across third-party logistics and mixed-SKU warehousing will still be unlikely. The economics and operational flexibility requirements won’t support it for most operations.
Is full automation an endpoint or an asymptote?
The evidence suggests treating full automation as an asymptote rather than a date on the calendar. Warehouses will get progressively more automated. The percentage of facilities with high automation will increase steadily. But the vision of the average warehouse running with near-zero human labor keeps receding as you approach it.
Consider what hasn’t changed despite decades of automation development: warehouses still need people for judgment calls, edge cases, and adaptability. The tasks requiring those human capabilities keep shifting, but they don’t disappear.
A Deloitte study from 2019 found that by 2025, 79% of companies expected human-driven processes amplified with digital tools. Only 18% believed operations would move toward lights-out automation. We’re now at that 2025 milestone, and the prediction held up. Most companies deployed automation to augment human workers, not replace them.
The pattern will likely continue. Technology will automate more tasks. Warehouses will need fewer people to process the same volume. But the remaining people will handle increasingly specialized and complex work that automation can’t yet address reliably.
So when will warehouses be fully automated?
If we define ‘fully automated’ as most warehouses operating with minimal human labor, the answer is: not in the next 10 to 20 years.
If you define it as some warehouses in specific industries operating at near lights-out levels, the answer is: it’s already happening in limited applications and will expand through the 2030s.
If you define it as your warehouse, the answer depends on what you’re handling, how variable your demand is, how standardized your products are, and what your labor and capital economics look like.
The research points to a future of islands of automation connected by software, managed by smaller but more skilled workforces. Warehouses in 2035 will look radically different from warehouses in 2025. But they’ll still have people in them, handling the complexity that automation can’t yet master.
That’s not a failure of automation. That’s recognition that warehouse operations involve more variation, more edge cases, and more judgment calls than most people realize. The technology will keep improving. Automation will keep spreading. Full automation across the industry will continue to be more aspiration than reality for the foreseeable future.
Sources:
Interact Analysis (2023). “Mobile robot market slows but robust growth continues”
Interact Analysis (2025). “Mobile Robot Market Forecast Slashed: Tariffs and Uncertainty Take Their Toll”
DC Velocity (May 4, 2023). “Warehouse automation on track to reach 26% of sites by 2027”
McKinsey Quarterly (July 8, 2016). “Where machines could replace humans and where they can’t (yet)”
McKinsey (April 24, 2019). “Automation in logistics: Big opportunity, bigger uncertainty”
MIT Center for Transportation and Logistics (2024). “The Warehouse of the Future”
MIT Center for Transportation and Logistics + Mecalux (December 2025). “The State of AI in Warehousing”
U.S. Bureau of Labor Statistics. “Warehousing and Storage: NAICS 493” (Industries at a Glance)
Gartner (2024). Cyber-physical automation adoption forecast
Deloitte (2019). Warehouse automation and digital transformation study
Mordor Intelligence (2025). “Warehouse Automation Market Size & Share Analysis”
Peerless Research Group for Modern Materials Handling (2025). Warehouse automation survey
MHI + Deloitte (2024). “Annual Industry Report: Supply Chain Innovation”
Financial Times (2025). Amazon capital expenditure and robotics investment reporting
