Humanoid robot or cobot: which fits your shop floor?

What is the difference between a humanoid robot, a cobot, an industrial robot and an AMR?

An industrial robot is a fixed robot arm built for speed and repeatability, usually behind guarding. A cobot is an industrial robot designed and safeguarded to share a workspace with people. AGVs and AMRs carry goods across the floor. A humanoid robot has a human shape, walks and handles objects with two hands.

The types often work side by side. In many plants an AMR brings totes to a cell where a cobot loads a machine.

• Industrial robot: an arm in a fixed robot cell, used for welding, painting or palletising. Designed for high speed, high precision and large volumes.
• Cobot (collaborative robot): usually a lighter arm that may work with people nearby, under the requirements of ISO 10218. Common jobs are machine tending, screwdriving, inspection and pick and place.
• AGV and AMR: mobile robots that move totes, carts and pallets. An AGV follows a fixed route along lines or markers; an AMR plans its own route with sensors and a map and drives around obstacles.
• Humanoid robot: a robot with a torso, arms, hands and legs, such as Figure 03 from Figure AI. It is meant for workplaces designed for people and for changing tasks that it learns.

Cobot vs industrial robot vs humanoid: how do they compare?

They differ in how freely they move, how close to people they may work and how mature the technology is. Industrial robots and cobots have been in common use for years and fall under existing ISO standards. For humanoid robots, a dedicated safety standard was still being drafted in October 2026.

Industrial robots remain the bulk of the market. According to the IFR, the number of industrial robots in operation in factories worldwide reached a record 5 million in 2025.

Industrial robot, cobot, AMR or AGV and humanoid robot compared, as of October 2026

Which safety standards apply to cobots and humanoid robots?

Industrial robots and cobots fall under ISO 10218: part 1 covers the robot itself, part 2 the application and the robot cell. The 2025 editions include the requirements for collaborative applications that used to sit in ISO/TS 15066. As of October 2026, no final ISO safety standard specifically for humanoid robots had been published.

A cobot is therefore safe only within an assessed application. A cobot holding a sharp blade or a heavy workpiece needs a different assessment from one that tightens screws. ISO/TS 15066 describes four ways of working together: safety-rated monitored stop, hand guiding, speed and separation monitoring, and power and force limiting.

Humanoids belong to a new category, because they keep their own balance and can fall over without power. ISO is drafting ISO 25785-1 for dynamically stable industrial mobile robots, which was at committee draft stage in October 2026. BMW reports that after its pilot with Figure 02 at Plant Spartanburg it revised the safety concept, with additional barriers and partitions.

Figure covers Figure 03 in soft textiles instead of hard machined parts, with multi-density foam at pinch points, and BMW lists these soft components as a safety feature. A risk assessment of the workplace is still needed.

Which type of robot is the most flexible?

A humanoid robot promises the most flexibility, because it works with two hands in spaces built for people and learns new tasks from data and demonstrations. An industrial robot is the least flexible and the fastest. Cobots and AMRs sit in between; you can reprogram them for a new task or route fairly quickly.

The International Federation of Robotics is cautious. Its 2025 position paper states that humanoids "will not compete with traditional industrial robots in terms of speed, precision, reliability and repeatability".

Figure runs Figure 03 on Helix, a Vision-Language-Action model that combines vision, language and movement. According to Figure, Helix uses one set of neural network weights for all behaviours, without fine-tuning per task. In a logistics task, Figure reports that going from 10 to 60 hours of demonstrations cut the average time per package from about 6.84 to 4.31 seconds.

Many cobots can be taught by moving the arm by hand into position. For changing products in small batches, that is usually the quickest route to a working application.

What does each type need from your workplace and systems?

An industrial robot needs a guarded cell and a steady supply of parts. With a cobot, the work is in the risk assessment of the whole application. AMRs need clear, level routes and a reliable network, while a humanoid robot needs room to walk, a charging spot, network coverage and, for now, often extra guarding.

BMW lists improved 5G coverage in the hall among the lessons from its Spartanburg pilot. Figure 03 charges wirelessly: coils in its feet charge at 2 kW once the robot steps onto a charging stand, so it can top up during the day.

Every robot needs instructions. An AMR has to know which tote goes to which line, a cobot which program belongs to which order. That information lives in your ERP, MES or WMS. The better those systems track what has to be made and moved, the easier a robot fits in, and that applies to a humanoid as much as to a cobot.

Which robot fits which task?

Start with the task. If one action repeats at high volume, an industrial robot is the obvious candidate. With an operator alongside, a cobot is more likely; for moving goods, an AMR or AGV. A humanoid robot comes into view for varied work you would rather not rebuild the workplace for.

The answer is often a combination. An AMR brings totes to a cell where a cobot loads a machine, and a humanoid robot might later take on the work between those fixed stations, such as the sorting BMW is testing with Figure 03.

Decision table: which robot for which situation

• Robots and humanoids in manufacturing
• Robots in logistics and the warehouse

When is a humanoid robot worth considering?

When tasks vary, take place in spaces built for people and are hard or costly to solve with fixed automation. Picking parts from a container and putting them in the right sequence is a typical example. As of October 2026, plan for a pilot: the technology moves fast and Figure publishes no price or order page.

BMW is the best-known example. At Plant Spartanburg, Figure 02 retrieved sheet-metal parts and positioned them for welding for almost a year. BMW reports that the robots supported the production of more than 30,000 BMW X3s, five days a week in ten-hour shifts. In June 2026 BMW announced a new use case there for Figure 03 in logistics sequencing: components arrive unsorted in larger containers and the robot picks them up and sorts them.

Figure says robots from its BotQ plant go to internal research, data collection and the development of commercial use cases, among other things. For most manufacturers and logistics companies a humanoid is something to follow and explore for now. A cobot or AMR you commission this year will usually pay off sooner.

• Figure 03 and what it does at work

How do you approach the choice?

Write down the task first: what is handled, how often, how heavy, where and next to whom. Then look at what the workplace and your systems need to arrange and which standards apply. Only after that do you compare robot types and suppliers. A small trial on one task tells you more than a brochure.

BeeManaged builds AI agents that work in the processes of manufacturing and engineering companies, connected to ERP, CRM and service systems. In an orientation we look with you at which tasks qualify, what that asks of workplace, safety, systems and people, and how a robot fits into the same process as your agents.

1. 01

   Describe the task

   Which action, how many times per shift, what weight and how much variation in products.

2. 02

   Look at the surroundings

   Floor, walkways, room for a cell or guarding, network and charging points.

3. 03

   Check your systems

   Where the instructions come from (ERP, MES or WMS) and who records what the robot did.

4. 04

   Compare the types

   Use the decision table above and the standards that apply to your application.

5. 05

   Start small

   One task in one place, with agreements beforehand on what the trial has to show.

• Humanoid robots in manufacturing and logistics
• Plan an orientation call