Kejie launches human-robot 5.0 warehouse system as robot picking hits 16.8%

Let me get the facts straight first, because there has been a lot of loose commentary about robots replacing warehouse workers wholesale. Kejie Logistics, the supply-chain arm of Digital China, launched its Human-Robot 5.0 system in Guangzhou on 22 September. It is built on a cloud brain plus edge brain architecture. The cloud brain does wave consolidation, pick priority and task dispatch. The edge robots handle identification, obstacle avoidance and picking. In the footwear and apparel warehouse run together with Dagai Robotics, the robot-picked share of orders went from 0.97 percent at launch to 16.8 percent within months. The stated goal is to move large-warehouse automation away from heavy, long-cycle construction projects and toward lightweight, replicable deployment.

Now why does this matter to an importer or exporter rather than just to a warehouse operator. Most of us who move goods across borders do not run our own DCs. We buy fulfillment, cross-border bonded warehousing and value-added services from providers, and we pay for them either as a fixed handling fee per order or as a variable labor component baked into the rate card. When a provider's automated picking share goes from near zero to a meaningful minority, the labor component of that rate card starts to change shape, and that shows up in your contract about a year later. I have been on both sides of this negotiation and the lag is fairly consistent.

The 16.8 percent number is worth unpacking because it is not the number people assume. It is not 16.8 percent of warehouse work, it is 16.8 percent of orders whose picking was done by robots. In a footwear and apparel operation the pick profile is unusually friendly to automation, because the SKUs are durable, relatively uniform in handling, and often stored in standardized tote-friendly units. Categories with fragile goods, high-value small items that need serialized scanning, or genuine multi-item gift assembly are much harder to automate, and I would expect their robot-picked share to sit well below 16.8 percent for a while yet. So if your product is electronics or cosmetics with heavy kitting, do not assume the number transfers, and do not let a provider quote you a headline facility number that quietly averages in the one category that automates well.

The cost side is where I want to do some arithmetic, because this is where the real picture sits. Assume a cross-border e-commerce bonded warehouse handles 30,000 orders a day and charges you a blended handling fee of 1.20 yuan per order, with roughly 55 percent of that being labor cost. That is 0.66 yuan per order of labor, or 19,800 yuan a day across the facility. Now assume robots take over 16.8 percent of that picking volume and the provider passes through, say, 40 percent of the labor saving, which is generous but not unusual when a provider has committed capex. The saving passed through is 19,800 times 0.168 times 0.40, which comes to about 1,331 yuan a day. Divided across 30,000 orders that is roughly 0.044 yuan per order, or about 3.7 percent off the handling fee. On a million orders a year that is 44,000 yuan, real but not transformative.

The reason I still take this seriously is the trajectory rather than the level. The system went from under 1 percent to 16.8 percent in months in one warehouse. If the architecture is genuinely lightweight and replicable, the next step is not 16.8 percent to 20 percent, it is 16.8 percent in one warehouse to 16.8 percent across many warehouses, and then the unit economics of the whole rate card shift. Providers that adopt early get to reprice before their competitors do, and they get to bid on accounts their competitors cannot serve profitably at the same fee. That is the strategic change, and it is why I would not treat this as a warehouse story.

There is a flip side that I want to be direct about. Automation changes the labor structure, and that is not a neutral thing for the service you get. A warehouse that has automated 16.8 percent of picking has moved its most repetitive work off human hands. What remains for humans is the harder work: exception handling, damaged goods, gift assembly, returns processing, quality checks and the odd orders that do not fit the standard flow. In my experience the people who stay in that warehouse are working on harder tasks with less slack, and the risk of a service failure moves from volume-driven to exception-driven. If you are an exporter relying on that warehouse to hit a carrier cutoff, a single mis-handled exception at 4 pm on a Friday hurts more than a slow pick at 11 am on a Tuesday.

Let me also put the onboarding cost on the table, because nobody ever budgets for it. Rushing a new automation system into a warehouse that serves your orders is not free. Barcode and label standards have to be re-verified against the robot's scanning hardware, tote and carton dimensions have to be re-standardized, the WMS integration has to be tested, and someone has to be accountable for the transition weeks when throughput dips. I have lived through a warehouse system cutover that cost more in the first month of service failures than it saved in the first year of labor. Assume two to three weeks of degraded throughput during a rollout, and build a buffer of at least 15 percent extra order capacity ahead of any peak season that overlaps with a go-live.

So what do you actually do with this. Begin by asking your warehouse and 3PL providers a direct question that most shippers never ask: what share of my orders is currently picked by robots, and what is the provider's plan and timeline for that share. Ask it in writing, ask for the number as a percentage of your own order volume rather than a facility average, and ask what happens to your rate card when the number moves. A provider that cannot answer is either not measuring it or not willing to share, and both are useful information before you renew. Push until you get an actual percentage tied to your order volume, and if the answer comes back as a facility average, ask for your own slice separately.

Next, build the rate card question into the renewal rather than tacking it on. Ask for a labor component and an automation component to be split out in the pricing, even if the headline number stays blended. Once the components are visible, you can ask for a mechanism that passes a defined share of labor savings back to you as the automated share rises. I would aim for a pass-through of 30 to 40 percent of verified labor savings above an agreed baseline, reviewed every two quarters. Below 30 percent the provider keeps too much; above 50 percent they will not sign and you will not get the transparency either.

After that, run a small live test instead of a big commitment. Pick one product category with a clean, automatable pick profile and route a defined slice of it, say 10 percent of your weekly order volume, through the provider's most automated facility for eight weeks. Measure three things religiously: on-time dispatch to the carrier cutoff, exception handling time, and the per-order cost delta. If the exception time gets worse by more than 20 percent even as cost improves, that is a warning sign and you should keep the automated facility on a portion of volume rather than moving everything there.

Then think about what this means for your own footprint. If third-party automated picking is getting cheaper and more replicable, the case for a fully self-operated warehouse gets weaker for all but your highest-volume, most standardized flows. I have watched several mid-sized exporters keep a self-operated DC out of habit and pay for it. Before your next lease renewal, run a genuine comparison between running the DC and buying the same service from an automated 3PL, and be honest about which of your product lines actually justifies owning the operation.

One more thing, and this is the one that catches people out. Robotics changes the physical handling standards your packaging has to meet, and those standards are enforced by the machine whether you agreed to them or not. If your cartons are within a couple of millimeters of a limit, or your labels sit slightly off-position, an automated pick line will reject them in a way a human never would. Get your packaging specs reviewed against the automated handling requirements at your key provider before the automated share goes above roughly a quarter of your volume, because retrofitting packaging under production pressure is expensive. I have seen a shipment held entirely because a barcode was printed two millimeters lower than the scanner window tolerated.

The bigger picture, and the reason I would not dismiss a 16.8 percent figure as too small to matter, is that warehouse automation in the China and cross-border supply chain has reached the point where it is a procurement decision rather than a technology bet. The technology works well enough at a modest scale. What is now being decided is who captures the savings, and that is decided in contract language, not in the press release. The providers know it. The question is whether the shippers asking them for quotes know it too. If you are buying warehousing for export flows, the next twelve months are when the pass-through terms get set for the next several years, and a shipper who asks the right three questions will end up on the right side of that split.

There is a piece of this story that gets almost no attention and it is the one I would push hardest on if I were sitting across from a provider. Automation capex is paid by the provider, and providers do not absorb capex out of kindness. They absorb it to win contracts, and they win those contracts because a shipper signed a volume commitment that makes the payback math work. The moment a warehouse commits to a robotic pick line, it needs predictable volume to hit its return. That need gives a shipper leverage at the signing table and takes it away afterward. If you are asked to sign a two or three year volume commitment to help fund an automation rollout, get something real in return, and I would define real as either a rate escalator capped below the published rate card, or a discount schedule that increases as the automated share grows, or an exit ramp that lets you leave after twelve months with a defined notice period. A volume commitment without one of those three is you funding somebody else's capital base.

Let me also address the category question properly, because I have watched shippers get this wrong. The 16.8 percent figure comes from footwear and apparel. That category automates well for specific physical reasons. Cartons are sturdy, units are relatively regular, and the pick is mostly a single-item task with predictable weight. Compare that with a category where your orders average four to six line items of small value, need serialized capture at the unit level, and get assembled into a gift presentation with tissue and a card. No current robotic pick line handles that gracefully, and the share you will realistically see in those flows is going to sit under five percent for a while. So before you build assumptions into a budget, look at your own order profile rather than the provider's showcase category. The right diagnostic is the average lines per order and the percentage of orders that are single-line, because a warehouse with 70 percent single-line orders is a very different automation candidate than one with 20 percent.

Now let me talk about what this does to your peak season planning, because that is where the money actually moves. Automated picking does its best work on high volume, predictable, standard orders, which is precisely the profile of a peak-season promotional push. If your provider has automated a meaningful share of picking before your peak, their peak capacity is structurally higher than it was last year, and that changes how much of your peak volume you can commit to them versus seek elsewhere. Assume your peak week is 3.5 times your baseline week, and assume the provider's automated share in your category reaches 20 percent by next peak. The incremental daily capacity from automation on a 30,000 order baseline is roughly 30,000 times 0.20 times the throughput multiple per robot versus a human, which in a well-run footwear pick operation might be in the range of two to two and a half times. That is a meaningful capacity addition, and it is capacity you should be negotiating for allocation on now, not in October when everyone else is asking.

The counter-risk is that providers sometimes chase automation to serve larger accounts and quietly de-prioritize smaller ones. If you are a mid-sized exporter, your peak allocation can get squeezed even as the provider's overall capacity grows. I have seen exactly that happen, where a facility's headline capacity went up 30 percent and a mid-tier shipper's peak allocation went down because the robots were dedicated to a bigger customer's flow. So watch your own allocation, not the provider's headline numbers, and get your peak slot commitment in writing with a named facility and a defined cutoff for order acceptance.

One more angle, and this one is about compliance rather than cost. Automated pick operations generate richer data than manual ones, and that data often includes serialized unit-level records, timestamps and operator identifiers. If your goods are subject to any kind of origin traceability, conformity assessment or recall regime, that data has value and it also creates obligations. Make sure your service agreement says clearly who owns that data, how long it is retained, and what you get if you leave. I have watched a shipper try to reconstruct a shipment history after a recall issue and discover that the warehouse's retention window was 90 days, which was 60 days shorter than the customer needed. That is a cheap clause to negotiate and an expensive one to discover late.

One last thing, and I will keep this short because I have said it in other pieces, do not let a vendor's technology announcement drive your decision. The relevant question is not whether the robots are impressive, it is whether the robots change your cost, your cutoff reliability or your peak capacity in a way you can measure. If you cannot measure it, do not pay for it. If you can measure it, put the measurement into the contract so it is not a matter of opinion later.

Author: Leo

• Ask every warehouse and 3PL provider by 31 October 2026 for the robot-picked share of your own order volume, their automation roadmap and the rate-card impact in writing.
• Negotiate a pass-through of 30 to 40 percent of verified labor savings above an agreed baseline, reviewed every two quarters, at every renewal from Q4 2026 onward.
• Run an eight-week pilot moving 10 percent of weekly volume through the provider's most automated facility, tracking on-time dispatch, exception handling time and cost per order.
• Build a 15 percent extra order-capacity buffer for any peak season that overlaps a warehouse automation go-live, assuming two to three weeks of degraded throughput.
• Have packaging specs re-verified against automated handling requirements before robot-picked share exceeds 25 percent of your volume.