Latest / The Manufacturing Podcast with Fexingo: Factories, Supply Chains, and Industrial Business / The Robot That Packs 1200 Boxes an Hour
Transcript
- Lucas: So we're at episode 100 of this show, and I wanted to do something tangible, something you could actually see on a factory floor right now. Luna: I'm ready. What'd you find? Lucas: There's a company in Denmark called RobotWorx — not a household name, but they make a robot they call Sloth. It picks and packs e-commerce orders. And the headline number is 1,200 boxes an hour. Luna: Wait — 1,200 boxes an hour? That's 20 a minute. That's one every three seconds. Lucas: Exactly. And to be clear, that's not just dropping items into a bin. Sloth has to pick the right item from a tote, scan it, place it into a shipping box, add dunnage if needed, tape the box, and label it. All in about three seconds. Luna: Okay, and how does it not crush things or rip the packaging? I've seen industrial robot arms — they're strong, not gentle. Lucas: That's the clever part. Sloth uses a combination of suction grippers and computer vision. The vision system identifies the item and its orientation, then the gripper applies just enough vacuum to lift it. The robot arm itself has force sensors — it can feel if it's pressing too hard. The engineers tuned it so it handles a bag of potato chips as carefully as a glass bottle. Luna: But it's called Sloth. That name feels intentional. Lucas: Very intentional. The CEO told me they deliberately made the arm move slowly — it tops out at about walking speed. They found that when robots move fast, human workers get nervous, stand further away, and the whole line slows down. Sloth moves at a pace that feels natural, so people are comfortable working nearby. And it's actually faster than a fast robot, because the system never stops. Luna: That's a really human-centered design choice. Most automation companies brag about speed in meters per second. They brag about perceived slowness. Lucas: Right. And it works. The robot runs three shifts a day, six days a week. They've had one installed at a Danish e-commerce fulfillment center since 2024, and in the first year it handled over 2 million orders with less than 0.1 percent error rate. Luna: What about the tricky stuff — like black shrink wrap? I've heard computer vision struggles with glossy or dark surfaces. Lucas: You're spot on. Black shrink wrap is actually Sloth's kryptonite. The vision system uses structured light — it projects a pattern onto the object and measures the distortion. But black surfaces absorb too much light, so the pattern doesn't show up. RobotWorx had to add an infrared camera just for dark items. Even then, they have a failure mode: if the system can't resolve the object, it signals a human operator to handle that one item. It happens maybe once every 200 picks. Luna: So it's not perfect, but it's good enough. And 99.5 percent success on 1,200 boxes an hour is still impressive. Lucas: Yeah. And the economics are pretty stark. The robot costs about $300,000 fully installed. One person can supervise three Sloth robots. So if you replace ten workers per shift, and you run three shifts, the labor savings add up to about $200,000 a year in a country like Denmark where warehouse wages are around $22 an hour. Payback period is roughly 18 months. Luna: And that's the part that makes people nervous. What happens to the ten workers? Lucas: The company that deployed it said they didn't lay anyone off. They reassigned workers to quality inspection and exception handling — the jobs that require human judgment. But that's one facility, one story. Across the industry, the math is hard to argue with. If you're a warehouse manager and your competitor installs Sloth, you either match it or your costs are higher. Luna: Which is why these conversations matter. If today was useful to you — maybe you're thinking about automation for your own operation, or just trying to understand what's coming — the way we keep this show ad-free and independent is through listener support. It's a small thing: buy me a coffee dot com slash fexingo. That's it. No tiers, no perks, just a way to say this kind of deep-dive is worth a coffee. Lucas: And it genuinely makes a difference. We don't run ads, we don't take sponsor money, so every contribution goes straight into research time and better episodes. We're grateful for every one. Anyway — back to Sloth. One detail I love is that the robot actually learns over time. Luna: Learning how? Lucas: Every time it fails to pick an item, it records the image and the gripper pressure. Engineers review those failures weekly and tweak the suction parameters. So after six months, the robot is noticeably better at handling odd-shaped items — like a bag of dog food with a weird seam, or a bottle with a slippery label. It's not true AI learning, but it's a human-in-the-loop improvement cycle. Luna: That feels more realistic than the hype about fully autonomous factories. You still need people to tune the system. Lucas: Exactly. And RobotWorx is very upfront about that. They say their goal isn't to replace humans — it's to handle the boring, repetitive picking so humans can do the interesting troubleshooting. But I'll be honest: I've heard that line from every automation company. The difference here is they designed the robot to work alongside people, not behind cages. Sloth doesn't need a safety fence because its speed and force are low enough to be safe around humans. Luna: That's a huge deal. Traditional industrial robots have to be fenced off. That takes up floor space and creates workflow barriers. Lucas: Right. And Sloth's footprint is about the size of a shipping pallet. You can drop it into an existing packing station with minimal reconfiguration. The company that installed it told me they had it running within two days of delivery. That's unheard of for industrial automation. Luna: So what's the catch? Why isn't every warehouse buying one? Lucas: Two big reasons. First, the upfront cost — $300,000 is a lot for a small business. RobotWorx offers a robot as a service model at about $8,000 a month, but that still requires a multi-year commitment. Second, the robot works best with a stable set of SKUs. If your product mix changes radically every week, the vision system needs retraining, and that takes time. Luna: So it's ideal for a mid-size e-commerce operation with a consistent catalog, not for a highly variable one. Lucas: Exactly. And that's the kind of nuance that gets lost in the 'robots are taking over' narrative. Sloth is a very specific tool for a very specific problem. It's not a general-purpose humanoid. It's a dumb, slow, expensive arm that's really good at one thing. Luna: And that one thing happens to be the most physically demanding job in many warehouses. Picking and packing is repetitive, it causes injuries, and turnover is high. If Sloth can take that over, maybe it's a net positive. Lucas: That's exactly the argument RobotWorx makes. They say their customers typically see a 40 percent reduction in workplace injuries in the first year, because workers aren't doing the twisting and lifting anymore. They're standing at a console, handling exceptions. Luna: I want to see one in person. Do they do tours? Lucas: They do. And I'm actually planning a visit next month. If I can get some audio, maybe we'll do a follow-up episode from the factory floor. Luna: That would be great. For now, I think the takeaway is that automation isn't one wave — it's a thousand small, specific changes. Sloth is one of them, and it's a well-designed one. Lucas: I'd agree. And as we hit episode 100, I'm glad we started the next hundred with something concrete. A robot named Sloth that moves slow to go fast. Luna: Here's to a hundred more — and maybe a Sloth sighting.