Are robots actually doing real work yet?
Yes, in carefully chosen jobs. Here are three results you can check.
AGIBOT
🦾 Electronics manufacturing · 📍 Nanchang, China
Case study 1: AgiBot G2 completed 65K factory tasks at Longcheer
production tasks completed
17,625
units of line output
AgiBot’s G2 robots worked inside Longcheer’s tablet-production workflow, handling tasks such as loading, unloading, testing and sorting. AgiBot reported a 99.99% task success rate across the run, making this probably the strongest publicly quantified Chinese humanoid-style factory case study today.
Boston Dynamics
🦾 Food & beverage · 📍 Leuven, Belgium
Case study 2: Spot found twice as many faults as expected
inspections a week, 10 lines
~150
faults found in 6 months
average repair time, down from months
Spot patrols packaging lines that produce 50,000+ containers an hour, using thermal and acoustic sensors. It caught a failing transport motor that would have cost at least six hours of downtime, found two CO₂ leaks within 20 minutes, and flagged gearboxes low on oil. A single ammonia leak can cost up to $15,000; expected payback was under a year.
Figure
🦾 Automotive · 📍 Spartanburg, USA
Case study 3: Figure 02 spent 11 months on a real production line
runtime on 10-hour shifts
BMW X3s built with its help
Figure's humanoids worked Monday to Friday, loading parts to within 5 mm in about two seconds each, inside an 84-second line cycle, with a placement accuracy target above 99%. They walked about 1.2 million steps (200+ miles).
Why it works: Humanoids and quadruped robots turn human labor into an engineering problem: work can be measured, programmed, tested and continuously improved. Instead of accepting labor shortages, fatigue, inconsistency or unsafe conditions as fixed constraints, companies can iterate on hardware, software and workflows until performance improves. Over time, labor becomes something that can be optimized much like any other production system.
Can robots actually work 24/7?
Increasingly, yes. With autonomous charging and battery-swapping systems, humanoid robots can now operate for close to 24 hours with only short interruptions. UBTECH’s Walker S2, for example, can swap its own battery and return to work without human intervention.
▶ Video: UBTECH Walker S2's battery swap guarantees near 24-7 uptime, at least in theory
But there are several catches. First, today’s robots achieve their highest reliability on narrow, repetitive, well-defined tasks. The more variability, dexterity and judgment a task requires, the harder it becomes to sustain high performance over long shifts.
▶ Video: Today, robots provide meaningful labor in highly repetitive and predictable workflows
Second, a robot does not operate in isolation. The surrounding production line, material flows, human workers and upstream/downstream processes all need to support the same operating pace. A robot may technically be capable of working around the clock, but that does not mean the factory around it can usefully keep it busy around the clock.
▶ Video: Figure 03 can continuously sort packages. But can humans that load the docks?
What jobs suit a robot today?
| Job |
Best robot type |
Proof it works |
Examples on RoboHours |
|
Moving totes, boxes and bins
|
Industrial humanoid |
GXO / Digit, Figure parcel sorting |
Agility Digit, UBTECH Walker S2, AGIBOT G2 |
|
Loading parts and tending machines
|
Industrial humanoid |
Figure at BMW |
UBTECH Walker S2, Agility Digit |
|
Inspection rounds (heat, leaks, gauges)
|
Industrial robot dog |
Spot at AB InBev |
Boston Dynamics Spot, Unitree B2 |
|
Night patrols and security checks
|
Industrial robot dog |
Route-based patrols |
Unitree B2, Boston Dynamics Spot |
|
Staff demos, investor tours, "what could this do here?"
|
Small humanoid |
Low-cost first step |
Unitree G1 |
Where do robots still fall short?
Be realistic. Gartner expects fewer than 20 companies to run humanoids in full production for manufacturing and supply chain by 2028. It notes they still cost more than task-specific machines, move slower and are limited by battery life. Other things to plan for on an industrial site:
-
Brains: true autonomy and generalizability of AI for robots is far from solved.
-
Speed: robots are usually slower than an experienced worker. Their value is consistency and covering hard-to-fill shifts.
-
Integration: linking a robot to your warehouse or production software takes set-up time. Ask who does it.
-
Safety: robots working near people need a risk assessment and clear zones.
-
Hardware wear: even BMW's trial found a weak spot (the forearm), which Figure then redesigned. Hardware generations are moving fast.
That last point is exactly why renting beats buying for a first deployment.See rent vs buy in our guide →
Why do rental prices look high for long-term arrangements?
Most robot rental quotes you'll see online, including many in our database, are short-term rates: a day or a few days, often with an operator on site, priced for events. An industrial deployment is a different deal. Under a monthly robots-as-a-service (RaaS) contract, the robot works many more hours, the operator isn't standing next to it all day, and the price per hour drops sharply. Agility Robotics, for example, says it targets a payback of under two years against a worker who costs $30 an hour fully loaded.
So use day rates to budget a short trial, not to judge what a year on the line would cost. Ask operators for a monthly or pilot quote.
Stay tuned! We're working with robot rental providers to bring longer-term rentals to market at a lower robo-hour rate.
How to Choose
Not sure which robot fits your site?
Our guide compares the five robot types, shows our top picks and gives you a two-week pilot template.
Compare robot types →
FAQ
Can a humanoid robot replace a worker today?
Not across a whole job. Today's robots handle specific, repetitive tasks such as moving totes, loading parts or sorting parcels, usually slower than a person but consistently. Most firms use them to cover hard-to-fill shifts and learn before scaling.
Are humanoid robots working in real factories?
Yes. Figure's humanoids worked 10-hour shifts at BMW's Spartanburg plant for 11 months, loading more than 90,000 parts for over 30,000 cars. GXO signed the first robots-as-a-service deal for Agility's Digit humanoids.
How much does it cost to rent an industrial robot?
Rental estimates vary by robot, location, dates, booking length and what is included. Browse the public catalogue for location-specific published starting estimates; they are estimates, not final prices. The rental operator confirms the final quote, availability, minimum booking and inclusions. Where a day estimate is shown, it is based on an 8-hour workday. Compare prices by city.
Should we rent or buy?
Rent first. Models improve every year, real use exposes weak spots (even BMW's trial found one), and a pilot shows whether the robot pays off on your task before you commit capital. See when buying makes sense.
Can a robot dog do inspections?
Yes. Industrial robot dogs such as Boston Dynamics Spot and Unitree B2 carry thermal cameras and acoustic sensors to spot overheating machines and leaks on a set route. At AB InBev's Leuven brewery, Spot runs 1,800 inspections a week, found about 150 faults in six months and cut average repair time to 13 days. See our top picks.
Can humanoid robots sort packages?
In live demos, yes. In May 2026 Figure live-streamed its Figure 03 humanoids sorting more than 100,000 parcels in 81 hours, about one every three seconds, according to company-reported figures. Real warehouse deployments are still mostly pilots, so test on your own parcels first. Use our pilot template.
How long should an industrial robot pilot last?
Two weeks on one repetitive task is a good start, measured against your manual baseline: tasks per hour, error rate, human interventions and uptime. Use our pilot template.
Do industrial robots need an operator?
During a pilot, usually. The operator sets up routes and tasks, supervises safety and handles problems. Some operators rent the robot only, which suits teams that already have trained staff. See what to check.
