Warehouse employment in the U.S. has doubled over the past decade as DCs ramped up hiring of hourly workers to pick, pack, and ship rising volumes of ecommerce orders. Investments in robots may slow the increase in labor demand in coming years, but robots are not the answer for all DCs. Most need a mix of strategies to keep labor costs in check.
Many DC operators are turning to new types of smart software that uses Artificial Intelligence to optimize labor productivity in non-automated processes. AI-based optimization attacks a major component of the labor productivity challenge by reducing travel. This is the same principal that underlies robotic picking systems, making AI a complement – and in many cases, an alternative - to robots for solving the DC labor challenge.
Labor Demand Will Continue To Grow, Despite Robots
According to data from the U.S. Bureau of Labor Statistics, warehousing employment in the U.S. more than doubled over the past decade. Most observers expect the trend to continue. This is adding to a difficult cost-calculus for many DCs, where rising wages and increased unit labor costs threaten to erode margins and profitability.
Robots alone will not stem the hiring boom for the foreseeable future. Amazon already employs as many as 200,000 mobile robots, but it is continuously adding to its warehouse workforce of about 900,000 full-time and part-time employees (excluding seasonal employees, as of September 2020). Although the pace of new hiring has slowed at Amazon, the reality is that DC employment by ecommerce operators will continue to grow, increasing labor demand and wages, especially in distribution-heavy areas.
Robots Are Not For Every DC
In a conventional, non-automated picking process, travel often represents the majority of the time in a DC associate’s day – ranging from 40-70 percent in most facilities. For years, DCs have used slotting, lean process initiatives, and conventional automation to reduce travel and boost productivity. Robots are the latest technology to address the challenge.
Advances in robotics over the past 10-15 years vastly improved the economics of goods-to-person systems that altogether eliminate travel in picking. Some providers have updated older AS/RS technology with more flexible, scalable robotic-type systems. Others have followed Amazon’s lead (following its acquisition of robot pioneer Kiva) by developing autonomous mobile robots (AMRs) that deliver products to workers at picking stations. Still others are using a robot-to-goods approach in which numerous robots roam in picking areas among workers, who walk between locations (typically in a single aisle) to pick and place the items in totes carried by robots. This strategy reduces, but does not eliminate travel.
Robotic picking systems are ideal for some operators, but they are not a fit for all DCs or for all products and customer types in every DC. For example, autonomous robots are generally not a fit for traditional grocery DCs supporting retail stores, or for B2B distributors serving a mix of retailers, wholesalers, and direct-to-consumer customers.
And while today’s AMR solutions offer advantages over earlier automation systems, they are by no means a low-cost solution. The systems require multiple robots per picker, and the cost for a single robot is about the same as the salary of a single worker. If a DC requires four robots per person to double picking productivity, the ROI stretches to more than eight years.
This is where AI-based software comes in, both as a complement to optimize robotic processes or as an alternative to robots for non-automated picking.
Using AI To Reduce Travel
In most DCs, order picking systems plan, organize and sequence picking tasks based on order priority, shipping cut-offs and other factors. They don’t typically consider product locations or proximity. Using AI-based software, DCs can evaluate proximity, location, and travel distances alongside priority and other factors to determine how best to organize, group (or batch) and sequence picks to minimize travel in the picking process.
The math behind this type of optimization is daunting. If 1000 orders are available for picking, and you are creating groups of four orders per picking assignment (or batch), there are more than 41 million possible combinations. AI-based tools can consider these combinations in milliseconds as users on the floor request work.
In item-picking processes common in ecommerce, the software can create optimal batches for picking items to carts, and create different batch sizes for different order types or cart configurations. Likewise, the algorithms can create optimal pallets of cases, or group together pre-defined pallets in two or three pallet assignments of work. Through this intelligent batching, DCs can slash travel distances in half, but that is only one part of the issue.
The second component of the travel challenge is to optimize pick paths or sequences. Most picking systems direct workers up and down aisles in a simple location sequence. AI-based tools can map out an ideal path that takes account of travel restrictions and other factors. Workers travel the shortest distance to complete their tasks, whether that is picking, replenishment, or another warehouse process.
Many ecommerce and multi-channel DCs can double productivity using AI-based travel optimization, depending on their pick process, batch sizes, and other factors. Grocery and other DCs that pick and ship cases on pallets might only see 20 percent productivity gains.
For most facilities, double- or triple-digit productivity gains are unheard of without radical changes and major investments in new infrastructure and automation. AI-based optimization doesn’t require any of that. It is a non-disruptive technology that can nevertheless transform warehouse operations and productivity.
