For a decade, piece picking has been the automation industry’s honest admission of failure. Pallets, cases, totes: all of that got automated years ago. But the single item pulled from a jumbled bin, in whatever shape, material, or orientation it happens to land in, remained the job robots couldn’t reliably do. Vendors kept promising the breakthrough was close. Operators kept hiring people to do it anyway. In 2026, that gap hasn’t closed entirely, but it’s closed enough that the conversation among logistics leaders has shifted from “is this technology real” to “which slice of our catalog can we actually justify automating this year.”
The Economics Finally Add Up, For Part of the Catalog
The number that’s changed the conversation is cost per pick. Manual picking still runs $0.35 to $0.55 depending on labor market, AMR-assisted picking brings that down to roughly $0.15 to $0.25, and fully robotic picking in structured environments is now approaching $0.08 to $0.12. That’s not a marginal improvement. It’s the difference between automation being a hard sell to the CFO and automation paying for itself in two to three years, with early adopters reporting close to 20% ROI within the first two years. The piece-picking robots market reflects that: it grew from $1.7 billion in 2025 to roughly $2.58 billion this year, and it’s forecast to reach nearly $20.8 billion by 2031, a 51.78% CAGR that isn’t the kind of number this industry throws around lightly.
None of that would matter without the labor backdrop pushing from the other side. Instawork’s 2025 State of the Warehouse Industry put workforce shortages as the single biggest challenge facing operators, with 76% of supply chain and logistics organizations reporting notable gaps. Piece picking is disproportionately exposed to that problem. It’s the most physically repetitive, highest-turnover role in most fulfillment operations. When the math on the robot side improves at the same time the labor side keeps deteriorating, the investment case stops being theoretical.
Who’s Actually Buying This, and Why It Matters
The deployments worth paying attention to aren’t pilots anymore. RightHand Robotics, one of the category’s more mature vendors, signed a multi-year agreement with Staples to deploy its RightPick system across fulfillment centers supporting next-day delivery to over 98% of the U.S. That’s a scale commitment, not a proof of concept. Closer to home, RightHand also struck an expansion agreement with Apotea, Sweden’s largest online pharmacy, to automate its new logistics center in Varberg, a useful signal that this isn’t only an Amazon-scale story; mid-size Nordic operators are making the same bet. And in March 2025, Rockwell Automation took a strategic stake in RightHand Robotics, aiming to fold picking directly into its controls ecosystem, the kind of move an industrial automation incumbent only makes when it believes the category has moved past experimental.
The competitive field has also settled into something more legible. Covariant and RightHand Robotics lead on AI-powered picking arms, Amazon’s in-house Sparrow system and Ocado’s robotics division represent the vertically integrated end of the spectrum, and Berkshire Grey and Plus One Robotics round out a market that, by most counts, now has around 4.7 million warehouse robots running across more than 50,000 facilities globally. That’s enough deployed volume that a leader evaluating this today isn’t betting on unproven technology. They’re deciding whether their own SKU mix and volume justify joining a market that already exists at scale.
The Hybrid Reality Nobody’s Marketing Slide Admits
Here’s the part vendors are less eager to lead with: full automation of piece picking, across an entire catalog, remains rare. The reason is the same one it’s always been. Warehouses hold a long tail of what the industry still calls “uglies”: items that are thin, deformable, transparent, shiny, or oddly shaped enough that vision systems and grippers still struggle with them reliably. That hasn’t gone away, and it isn’t going away this year.
What’s changed is that operators have stopped treating this as a reason to wait and started treating it as a segmentation problem. The practical 2026 model isn’t “robot vs. human,” it’s robots handling the 70 to 85% of SKUs that are structured and high-volume enough to automate profitably, with people handling the exceptions: the fragile, the irregular, the low-volume long tail that would cost more to automate than to staff. That’s a fundamentally different question for a leadership team to answer than “should we automate piece picking.” It’s “which segment of our catalog has the volume and uniformity to clear the ROI bar today, and which segment doesn’t yet, and won’t for another cycle or two.”
The Decision Isn’t the Robot, It’s the Fit
This is where the piece-picking conversation runs into the same mistake I’ve written about with WMS rollouts: the failure mode isn’t choosing the wrong vendor, it’s buying capability that doesn’t match the operation’s actual order profile. A robotic picking cell sized for a high-SKU, low-volume specialty retailer solves a different problem than one sized for a pharmacy fulfillment center moving high volumes of a narrower, more uniform catalog, which is exactly why Apotea and Staples are automating the same underlying task with different configurations. The question worth putting to any vendor isn’t “how fast does it pick,” it’s “show me your mis-pick rate and cycle time on our SKU distribution, not your reference customer’s.”
The financing question matters just as much. Robotics-as-a-Service has become the default entry point for the same reason it did in broader warehouse automation: it turns a large, hard-to-reverse capital commitment into a scalable operating cost, which matters enormously when you’re only confident about automating part of the catalog and want room to expand the program as the technology, and your comfort with it, improves. Buying outright still makes sense at the high end of volume and certainty. Renting makes sense everywhere else, which for most operators in 2026 is most of the catalog.
The Common Thread
Piece picking was always going to be the last mile of warehouse automation, and 2026 is the year it stopped being a research problem and became a segmentation-and-financing problem instead. The economics now clearly work for the structured, high-volume slice of most catalogs. The cost-per-pick numbers and the vendor commitments from Staples to Apotea to Rockwell make that hard to argue with. What hasn’t changed is that the long tail of awkward, fragile, low-volume items still needs a human hand, and pretending otherwise is how automation programs overspend and underdeliver. The leaders getting this right in 2026 aren’t the ones chasing 100% automation. They’re the ones who’ve done the unglamorous work of mapping their own SKU profile against what the technology can actually do today, and buying, or renting, exactly that much.
