|  | | | | ON INDUSTRIAL ROBOTICS The next wave of the robot revolution
| | | | | | | | | | | |
| Robots have been a fixture in industrial settings for the past 50 years. But until recently, breakthrough innovation wasn’t a hallmark of the sector. For the most part, preprogrammed robots performed repetitive tasks such as welding or painting, and they were not programmed to deal with disruptions. They couldn’t analyze a situation and adjust their actions to adapt. However, that’s changed over the past three to five years, thanks to technology improvements that have led to the creation of general-purpose robots that can perform multiple functions in different settings. No single technology served as the big unlock that enabled this. Related, complementary technologies—including transformer architecture, foundation models, generative AI, and actuators (the robot parts that execute movements)—all advanced simultaneously.
When I talk to clients, there’s a bimodal distribution of excitement about robotics. Some people, particularly those who have seen demonstration videos, are extremely excited and think that robots ought to be part of an immediate plan. Others are skeptical and think that salvation is still 20 or 30 years away. My view is that people will be surprised by the pace of innovation over the next few years. The first humanoid platforms—robots that are roughly the size and shape of humans and able to fit into spaces and contexts designed for humans—took several decades to assemble. Then, in the mid-2010s, a few companies started requiring only four to five years to develop original humanoid platforms. And newer companies, which have benefited from recent advances, have been able to build platforms in under a year.
We’re going to see big changes in the types of robots that are used in the workplace. Today, most robots are still fixed in place, with industrial arms that can move on up to six axes. The other big category, particularly in logistics, is mobile robots—think of them as large robotic vacuum cleaners, but instead of vacuuming, they carry objects. Those two classes account for the vast majority of robots in industrial settings.
It’s difficult to estimate how that breakdown will change in the future, but I think the first wave of general-purpose robots will be mobile manipulators that have wheelbases with arms on top. Bipedal robots could also perform a reasonable share of tasks. These robots may not be entirely humanoid, but they will have two legs that allow them to move. True humanoid robots could then be deployed commercially at scale, but that could take months or years to fully realize.
| | |
| | | “My view is that people will be surprised by the pace of innovation over the next few years.” | | | | |
| Not only will the number and types of workplace robots increase, but so will the degree of collaboration. In the past, many robots were siloed away from humans (sometimes literally kept in cages), but general-purpose robots will be integrated into the workplace. Robots will also be deployed in different but equally exciting ways across more sectors. For instance, some companies are using robots in hospitals to move documents, tools, and equipment needed for operations, which frees nurses to spend more time with patients. Robots are also being used in agriculture, not only for picking crops, but also for tasks such as weed killing.
Greater use of robots will naturally heighten safety concerns. Some relate to machine safety, such as how robots will respond to failure modes (including a lack of connectivity or power) when they are mixing with humans. No one wants a humanoid robot toppling over the moment it loses power, because that would be a hazard for anyone nearby. Other concerns relate to cybersecurity. Robots are connected devices, and we must ensure that they can’t be hacked and deployed in dangerous ways.
Beyond safety, companies will be addressing many other hardware- and software-related challenges over the next few years. Think about the difficulty of replicating a human hand. Our fingers can compress and adjust when we hold objects. Robot fingers are stiff and rigid, so they lose a degree of nuance in terms of sensing and manipulating.
Training robots is another issue. We don’t yet have enough data for the foundation models that allow robots to “think,” so researchers are performing simulations that generate synthetic data. All companies should be thinking about their data strategy right now. Think of companies that, for instance, assemble products using wrenches and other tools. The data that’s available to train robots today might include a video or written instructions explaining how to assemble a product. What’s missing is, for instance, a measure of how much force or torque workers must apply when screwing bolts. Companies could get that data by putting sensors on the wrench.
While the future could take many twists, one thing is certain: Companies that want to use robots must begin undertaking change management strategies and thinking about integrating robots within the IT stack. Do they have workers who can program robots? Or repair and maintain them? It’s best to consider these issues now, since the battle for talent will only get more intense.
| | | —Edited by Eileen Hannigan, senior editor, Boston | | |
| Share Ani Kelkar’s insights
|
|
|
|
|
| | | | Ani Kelkar is a partner in McKinsey’s Boston office. | | |
| | | | |
|
|
|
Copyright © 2025 | McKinsey & Company, 3 World Trade Center, 175 Greenwich Street, New York, NY 10007
|
|
|
|
|