Walmart spent billions replacing warehouse workers with robots, only to discover the robots can’t lift dog food and the engineers cost more than the workers did. Apparently the automation revolution has reached its greatest technical challenge yet: doing the job it was supposed to automate. Somewhere, a PowerPoint deck is quietly being deleted.

Robots can work well in warehouses. Amazon’s have been a success, for example. This just sounds like whomever WalMart hired to design this didn’t know what they were doing and Walmart got screwed. Maybe they tried to automate the wrong tasks, or leaned too heavily on the sales pitch of “AI magic” rather than understanding what they were buying.

The article also mentions Kroger’s failed dive into automation. From what I can find online, the automation worked, but it was a flop because it was expensive and underutilized — it wasn’t worth the money spent because it was focused on filling online orders that dried up.

wsj.com/…/amazon-warehouse-robots-automation-942b…

…substack.com/…/krogers-350-million-lesson-why-bi…

macrohint.com/krogers-2-6-billion-warehouse-meltd…

Ocado! I remember watching a promotional video about this being the future of grocery stores like a decade ago, and always vaguely wondered what happened to it. The whole hive-picker robot-carts-on-rails thing seemed very well developed, but I wasn’t convinced it was really practical.

Thanks for closing that loop for me.

I work at a Kroger owned store. For whatever reason they decided to give that another try at my store and another one in the same division. We’re all very sceptical… since you know it already failed before but it’s better than the robot “Tally” which drives around and checks for empty spots on shelves. The poor grocery guys are supposed to then go around and face em despite already being tight on people

The same items at my local Ralphs (Kroger) are constantly (like 4+ years) out of stock because they just face over the holes. Whoever is putting inventory into the computer must absolutely suck too because this shouldn’t even be happening; but having worked at Walmart I’ve seen firsthand how people roll their eyes when the numbers are off.

I’ve been told that it’s mostly automated. IE the Point Of Sale tells the system what items were sold and it subtracts from whatever total it has. The produce guys have said that because of this system some things like bananas are almost always screwed up, cause the barcode and the PLU (the numbers) aren’t actually the same item. Namely sometimes when the banana suppler changes they don’t get automatically ordered cause the system thinks we’re selling the old ones which “don’t exist anymore”. I’d expect similar things happen to grocery and home but to a much less extent, probably from people just tossing stuff out instead of scanning it out and then still just tossing it. Yay food waste.

They do quarter, semi annually, or anual inventories to true up the numbers

The system designers get tripped up by the strangest things. I once worked in a factory that uses copper bars of varying lengths, and the designer didn’t expect the bars to ‘sag’ when picked up flatwise from the middle. So I can imagine picking up a bag of dog food would do that.

That’s a bad engineer problem. Tasks must be reduced to absolute minimums, and operating scenarios must be maximized. Otherwise failure conditions will win.

They might have underestimated the mix of items. From bagged dog food with 20kg, boxes of all shapes down to small poly‘s is hard to handle. For picking robots as well as the sorting AGV‘s.

Amazon didn’t robotize its entire item spectrum yet, as far as I know. It focus on robotizable items.

If such a heavy dog food bag brakes, all the small pieces lie on the floor that those dumb AGV‘s drive upon. That’s immediately causing clogged wheels and even more. And to clean the mess up, one has to stop the entire AGV area as it’s not safe for humans to enter. One broken dog food bag can cause some expensive maintenance and drives operational costs up. 1 broken bag per month can likely kill your business case.

They just need robots that will clean up the dog food.

Simple thoughts for a complex problem never lead to a solution.

You might think one or two steps further. The 5-W framework might guide your thinking.

And here I thought the /s was loud enough.

Yeah just throw a few roombas out there and they’d be good to go. 👍

I mean anyone who’s ever done anything with a long piece of lumber or piping or anything could predict that.

You can bet your tushy that whichever middle-manager came up with the idea of spending all that money to reduce reliance on human workers, has already picked up a hefty bonus, used a golden parachute, and is now scamming their way through another giant corporate to repeat the same shit with a different idiotic idea.

Middle manager? I’m betting this idea comes straight from the top.

Nah, the top (C-suite) rarely has original ideas. Their job isn't to innovate or improve but to maintain the status quo. And they wouldn't risk their position with an idea that can backfire so easily.

No, that's reserved to the middle managers who have something to strive for - more money, more influence etc.

Since when is automating jobs an original idea? Middle managers absolutely don’t make decisions this big. Some bigwig heard about it from one of his buddies and announced a “transformation” a couple days later.

Automation isn't an original idea. This specific way of implementing it is. And it definitely came from a risk-taker, which would be a lower than C-suite manager.

It’s one of the great ironies of our time that we spend billions on slop machines but haven’t yet been able to build a robot that can empty the dishwasher.

As a bachelor, a dishwasher is already a cabinet. It doesn’t need emptying.

I’ve considered buying a second dishwasher so I can just swap between two self-cleaning cabinets.

So instead of hiring actual people again, they’ll look for cheaper engineers and robots.

Nah it’s at least another decade of sunk cost management style then they will pivot back to humans once minimum wage has been repealed and the labor camps are in place

they tried this like 10 years ago already.

So much of automation is trying to put robots in places humans were ment to be instead of designing processes that are for robots not people.

It’s no wonder to me there is often issues

I’ll never understand takes like this. You think humans are meant to be shuffling around a warehouse picking and packing orders mindlessly?

I get being annoyed at corporations for greedily trying to save money at the expense of workers, but ultimately, I can’t be that upset over shit jobs going away.

In my view, we have to be focused on how to help The displaced workers which we have historically done a terrible job of. It only becomes more important as we are able to automate more and more tasks that previously required a person. Automation absolutely can be a tide that raises all boats; we just have to demand that it be so. Otherwise, all the benefit will go to the oligarchs instead of the people.

The process is designed around human ability, not what is easiest to implement in a machine.

Our capabilities are different. I don’t mean that humans are biological geared towards working on a warehouse. But that we design warehouses around what humans can do

No, they aren’t saying that “humans are meant to be shuffling around a warehouse picking and packing orders mindlessly.”

The point is that the processes are designed around human operators with specific considerations for the abilities and constraints of an average human. Dropping a robot into a process without updating said process around the abilities and constraints of the robot creates new problems. Humans and robots work in different ways and cannot be substituted 1:1 without creating problems.

What’s interesting is that this specific company isn’t doing that.

They don’t build robots to move around a human warehouse, they build a giant warehouse (sort of) designed for their robots.

I work in this field and can speak to these systems.

These systems tend to perform quite well as their primary selling point is not necessarily cost reduction, flat, but rather increased revenue through higher and more reliable throughput. The trade off is that these systems are quite complex and normally require a great deal of coordination between the customer who may or may not operate the product and the OEM that builds and maintains it.

In my experience the biggest challenge is defining those responsibilities such that throughput is maintained. These systems are like high performance sports cars, they work well if taken care of and operated skillfully but can degrade faster as a result of lack of experience. Some of our customers permit us (the OEM) to operate and maintain to a greater extent then others. Given we have a greater deal of experience with the tool those customers tend to have measurably better performance.

To give an idea of why this particular industry has a harder time building the talent pool I’m referring to consider the domains that need to be covered.

Electromechanical systems, mobile robots, multi-robot systems, PLC programming, DCS, SCADA, multi-axis robots, VFDs, electrical power systems, relational databases, non-relational databases, message brokers, caching systems, distributed systems, Linux and Windows system administration, networking, industrial networking protocols (it’s different), software development, Kubernetes. This is the skill set of one technical professional in this space.

I commonly deal with issues that cover several of those areas simultaneously. E.g. robot communications are slowing down due to full message broker queue which is determined to be caused by a faulty motor causing slippage which leads to a large number of messaged being generated that fill the queue and cause other bots to begin making pathing errors due to communication timeouts.

So. The usual “Management trying to save a insignificant amount by making a stupid decision on a system they don’t understand for a job they don’t know”…

Ummm, no?

The system is way more expensive than the workers were. By like, orders of magnitude. They buy it because you can basically 10x a warehouses throughput using it.

Don’t get me wrong, after accounting for increased throughput and the gains by automation you end up with a bigger margin than by having a swarm of workers in a dusty warehouse but it necessitates hiring a ton of technical professionals that you generally have to pay more. Additionally this whole space is really hard for AI to touch because it’s very cross domain and real time. For reference. My role is considered “entry level” and I get 6 figures.

All that being said this isn’t so much managers being idiots (though there’s plenty of that) so much as “our core competencies built from the ways we did warehouses before do not translate to how we do them now and friction is being generated as a result of that transition.”

There are technical issues with the system. It’s largely built to integrate with legacy industrial tech which ties it into the windows space, this has led to a lot of decisions about tooling that degrades performance and leads to memory leaks.

Sounds like they might be learning from their mistakes though? If so, they’re smarter than most businesses.

The industry is still in its infancy, not just warehouse automation but robotics and fleetops in general.

There’s a lot of growing pains as we figure out the best ways to manipulate autonomous robot swarms that can number in the thousands.

One of the big issues is just training personnel for all this. It’s quite rare to find people with a background in both hardware and software and who can speak both fluently.

For reference. My role is considered “entry level” and I get 6 figures.

Please give me a job, thank you good sir or madam.

This is actually a good time to get into the space. There’s not a ton of companies competing over a now growing market and the work is weird enough that AI hasn’t fully eaten it.

Downside, see my other posts, the field wants to pull people who have a pretty diverse skill set. I would be prepared to discuss RAFT consensus in the same conversation as PID tuning and control theory.

Checkout companies like Symbotic, Boston Dynamics, or robotics integrators like ABB

Job sounds interesting. Several questions if u don’t mind me asking because I’m thinking to do small career change.

• What’s the job title I should look for?
• Do you think it’s possible to transfer from software developer? I have engineering background so I think I should be able to pick things up
• Is ageism real in your industry? I’m in 30s so ain’t young so worried about applying to entry level positions

Unfortunately we don’t have standardized roles in this industry.

Jobs like these tend to touch one of a couple spaces.

1. Controls engineering: PLC programming and traditional robotics
2. Fleetops, application engineer, or operations engineer: I’ve seen these show up at places like Boston dynamics
3. Technical support: look for ones that mention wanting like a EE and knowing stuff about control systems.
4. SCADA engineer: Especially at tech startups.

In general you’re looking for jobs that reference a mixture of networking, control systems, SCADA, DCS, PLC, IT infra, and of course robotics.

My entire team is 30+, hard to find people with a diverse background that are also 22.

These systems are MASSIVELY complicated and require quite a bit of technical know-how.

You named one main problem here.

The other ist cost. How much does a system cost that replaces one worker cost? INCLUDING the required technically skilled service staff?

As I replied to others, in systems like this it’s not as simple as cost savings. The selling point is increased throughput. The system costs more upfront and factoring future gains from increased throughput you end up making more money than if you had kept the cheaper manual workforce.

While I know you wanna stand up for your employer: don’t.

Quite literally for your job safety you should delete this post unless you’re literally PR and speaking in a way your employer would be fine with.

Most corporate training modules involve clear information you should never post like this because you’ll almost certainly get fired.

If you want to share things, post it more as “I work for a company that makes robots like this. The problem with our biggest warehouse clients is X, where our smaller ones is Y, so that’s probably the case here too.” Never directly show your employer like this.

Username says they run rum. I think the robot stuff is just a hobby.

Probably right lol, edited to remove identifying information.

Ah yes, the gotta get your SLA to PMS the PLC on your ABBs

You get what I’m putting down

So I’m all for these kinds of systems. What I see coming from a legacy manufacturer is they have so many limitations and restrictions, or their CTO and similar departments refuse to accept some of the system requirements, that once they purchase a novel new technology system, it just falls apart. From my experience it has almost always been two things, either the vendor selling the service over sold their ability, or the purchasing company ham strung the project so much that it was doomed before it started.

But I’m still hopeful, you actually build a system that meets the requirements, you have a customer actually embrace the systems and changes, it would enable success.

I’ve seen the same. My current take is basically “nothing about how we did warehouses transfers to this new way, and attempting to make them dovetail gives you the worst of two worlds”.

The cool thing is there’s so many neat new industries this can create. The downside is that, as usual, there’s a big disconnect between P&L decision makers and technical professionals.

Disconnect between P&L decision makers and technical professionals, been the same story for 40 years now. The more the C suite is filled with “business/accounting/law” the further the business moves away from its core competency. Praise be to jack welch, may he rest in piss.

Robots are more expensive and are less capable than people? What a surprise! Not. I saw that coming years ago… 🙄

Wait, you don’t know all the facts yet! We have a free trial for you of our fully automated hairdresser! You will be stunned how a whirlwind of blades trims the stuff growing out of your head in a mere minute! Without boring human conversations! For a little fee of 200 dollar per trail!

We had our super-intelligent robot hairdresser clone the voice of your old human hairdresser. It knows all the gossip, remembers the names of your grandchildren and can even tell you jokes and riddles! You should feel safe spilling your secrets to Carina the RoboBarber.

Unfortunately the robots are substantially faster and more effective than humans. Can you run at ~45 mph?

The issue is complexity and that’s something that’s getting better over time.

sprinkle in a bit of AI into the mix and now it’s not only more expensive, but also highly regressive.

My stand up Jack can go just as fast. The problem isn’t top speed, it’s acceleration and breaking. It’s being able to rip around a warehouse building orders quickly and efficiently. A dude with a stand up Jack and a voxcoder can beat most robots.

The company this post is referring to doesn’t have bots moving around a warehouse designed for humans. They built a whole building with shelves and space designed for the robots.

I promise you, the throughout is tremendously higher than humans alone. The bots can get up to roughly 45mph in a few seconds and break to a full stop in the same amount of time. I know this because I work in this space, the system is fast

And it actually works? They can’t even figure out how to make vending machines that don’t cause a log-jam.

I work for warehouse automation too. It absolutely works. Its just very carefully, meticulously planned. Every tiny edge case is accounted for. Hell. We measure the floor for difference in height to the mm where 3 is a big issue.

Thats kind of the whole problem with robots as tools. They can be incredibly fast and precise but if something is just slightly out of expected alignment they tear themselves apart (or kill someone nearby).

Same could be said for using a saw, or electricity.

If its designed correctly, conforms to safety requirements, there is no real risk to human health.

Of course a not can always crash, but that should be accounted for and tested against much like go carts on a track

Yep it works. When it’s running one of these things can push, not joking, millions of cases through a single warehouse a day.

Ever notice that the top end speed of the non riding ones, the ones that you walk behind, us just a little beyond a comfortable walking pace? Always found that curious. Not quite running, but certainly not walking either. Hips would hurt at the end of a shift.

Unfortunately robots are not able to think critically

They don’t need to. No more than a conveyor needs to think critically.

I think people are envisioning the robots moving through a warehouse that was built for humans. That is not the case.

Special purpose warehouses are built with the robots in mind. The bots don’t need to navigate an uncontrolled environment as the environment was meant for them.

It’s basically a really big, really performant conveyor belt.

Complexity is expensive.

Expensive isn’t really an issue though is it?

If I can guarantee you quadruple your revenue by doubling your expense, you’d take the deal no? That’s basically what these things do.

To be a bit more precise, is it about throughput or lead time? Because for throughput I imagine you can just have more workers, but for quicker shipping you need to do pick and pack in less time, rather than do a lot of them in a given time.

Throughput primarily, the bottleneck is still shipping. These systems do allow for picking/packing faster as well but from a logistics standpoint if you want more output you just need more trucks.

How does it help in getting more revenue then? Demand wouldn’t change, right?

Businesses tend to bank on demand for products being functionally infinite over the long term. Because of this it tends to be a safe bet that you’ll make more money the quicker, and cheaper, you can get products on shelves.

This has held pretty true thus far, though personally I have to wonder if there’s an upper limit. Human’s have to have a upper tolerance for buying junk right?

In this scenario, as you said it’s not making the delivery quicker, is it right to assume that it’s making it cheaper? Is the cost of warehousing lower with robots?

Especially at Walmart when loading a truck means throwing the shit inside and as fast as possible. Used to unload Walmart trucks. Wish had pics (this was wat before smartphones) You open back doors and the boxes would just spill out. There was always breakage. But Walmart didn’t care. It fuck up way Walmart does business and they are shady as fuck. But allows them to not care about the products. But yeah robots are slow and no way can work at the speeds required at a Walmart distribution facility.

There’s lot of types of warehouse automation outside of a robot skating on the ground picking up pallets or cases or anything. As in, there are certainly automation systems that work alongside people or other different bots that absolutely thrash the throughput of a typical warehouse operation

Walmart merely made a number of very stupid decisions when entering the automation game, but to them its drops in a bucket. They’ll just move on to the next one

unskilled warehouse workers.

It’s ironic to call them unskilled in an article describing how their skills are beyond billion-dollar robots.

“Unskilled” meaning you can teach someone to do the job reasonably well off the street in a few hours. Unfortunately robots are not humans and thus need to be meticulously designed while a guy just needs a decent paycheck and benefits.

There is no such thing as unskilled labor. All labor requires myriad skills. At best, some jobs are easier than others, for some people.

All references to “skilled” labor are somber reminders of centuries worth of capitalist propaganda implemented solely to underpay workers.

The length of training required for a job is irrelevant. Some jobs may require years of training for a repeatable mechanical task, like watchmaking. Others may let the worker pop off on day one with virtually no guardrails or training, like landscaping. Both jobs are difficult, and both workers deserve respect.

Do not call work skilled or unskilled, because it is not trueand such labels accomplish nothing other than serving bourgeoise interests.

100% agree. I work in the nuclear sector and jobs can literally be walk around and make sure nothing is on fire but its considered “skilled labor.”

Unskilled means can get someone to do it for minimum wage 🫢

It doesn’t mean it’s only worth minimum wage to the company.

Unskilled, but none of the managers I had (not WM, mid size mail order company) could keep up. All of them had broken racking, run the forks through walls, and/or torn dock doors off. They were slow, and would put bad pallets in the racks.

So the company liked doing cutting edge corporate stuff and reading books about various CEOs. One time they had us take skill surveys (thinly veiled IQ tests) to optimize new hires. The first batch was based off the high performance people and all quit after two weeks because they were smart enough to know bullshit. The second group was based off the long timers and also didn’t last because they were dumb as rocks. After that they went back to shotgunning everyone at it and seeing what sticks. Open interviews, always posted.

Remember when the disclaimer “please allow 6 to 8 weeks for delivery” was practically ubiquitous on mail order packages? It’s almost like that was a SANE way to run a logistics network!

Worked at a place that got obsessed with same or next day shipping. It was clearly killing employee morale and we weren’t selling anything that people needed that quickly. Make a special premium instant ship or something for the people having an emergency but not for everyone. The killer for me was that they were at times doing mandatory overtime during the week and then sending you home early on your required weekend shift so you could have the suck of getting your evenings and schedules killed but without the extra pay.

This is kind of what happened to me recently. Ironically I was fixing espresso machines. Between people expecting next day day of shipping to me being pushed to fix things faster and faster. I gave up and said fuck it. I probably did $50-70K in lost sales by not fixing the machines there and just quitting.

We don’t send pieces of paper back and forth with days long transit time to complete an order anymore.

Now displaced warehouse workers learn to maintain or program the bots and adapt to new jobs, but at some point, all this bot productivity should be taxed too or the wealth keeps concentrating among the few, rather than revenue to build better society for We The People.

Just tax wealth hoarding, doesn’t matter how the hoarding was achieved

That requires affordable education. Which is against the policies of those wanting you dumb so you continue helping vote for our one-sided policies that do nothing to help self but only the needs of an already overabundance and selfish select few.

As a forklift monkey: I hate my job and feel that humans weren’t meant for this monotonous kind of work. Like boomers who screwed in one nut on an assembly line who made enough to provide a family of 4, I’d blow my brains out before I’d do that. So fuck these jobs.

Yeah, the problem isn’t automation, it’s billionaires pocketing all of the proceeds.

Link to the original article. Oh wait it’s WSJ slop.

Anyone who has tried to automate anything beyond the most basic step quickly finds out how difficult it is to truly do it with all of the boundary conditions, bad inputs, breaking conditions, etc.

“We’ll pay billions in trying to automate things to avoid paying a few million for human employees.”

Walmart is technically competent, but more pragmatic than adventurous. In the past, it has often favored proven systems that deliver reliable returns over expensive, leading-edge alternatives. That has been a smart business strategy for them for the most part, even if they are sometimes late to the party when it comes to innovation.

Amazon has made robotics a core part of its fulfillment model. Ocado designed its grocery fulfillment operation around automation from the outset. This meant major investment in infrastructure, software, and process redesign.

Walmart has invested heavily in robotics, too, but its approach has been more incremental and constrained by the need to modernize an enormous existing network. A bolt-on, if you will.

So I suspect Walmart’s problems are partly about its approach and baggage, not the state of robotics. They’ll figure it out eventually.