Automation in Manufacturing Operations

Manufacturers are expecting to more than double their use of automation, artificial intelligence and other advanced technologies by 2030, according to an industry outlook from PricewaterhouseCoopers published in February.

The U.K.-based professional services firm, also known as PwC, surveyed 443 industrial manufacturing executives around the world for the report. The median share of respondents believe that advanced technology adoption throughout their operations is set to increase from 26% to 68% over the course of five years.

Two areas with the heaviest use of advanced technology today are production/operations and product design/development, according to the report. Other areas where adoption is relatively low, such as business support functions like finance and human resources, are set to quadruple by the end of the decade.

The race is on for manufacturers looking to enable AI and automation across their operations to boost efficiencies or create growth opportunities across the $16 trillion sector, but not everyone is starting from the same place. 

There is a widening gap between the “future-fit” companies and those falling behind due to poor data quality, skills gaps or fragmented systems, Ryan Hawk, global and U.S. industrials and services leader at PwC and author of the manufacturing outlook, said in an email. The gap could grow further as technology and capability advantages build on one another.

“The question is no longer whether companies will adopt new technologies, but how fast they can integrate them,” Hawk said. “As automation becomes ubiquitous, the advantage shifts from who has tools to who can orchestrate them across the enterprise.”

Automation of key business processes, such as data capture and analytics and physical production, is set to boom. According to PwC, adoption of highly automated processes will nearly triple by 2030, extending from the front and back offices to research and development and the shop floor. Survey results show a more even distribution of technology enablement across the value chain, not just in production/operations.

Hawk said this even spread was “most eye-opening.” It signals a fundamental shift toward AI-supported engineering, digital twins and more simulation-driven development, “not just smarter factories,” he said. There’s also growing interest in sensor technologies and physical AI.

As technology enablement and automation surge, manufacturers are looking beyond their core products for ways to generate revenue. They are positioning themselves as providers of “integrated solutions” that combine hardware with software, data, services and technology support, Hawk said. Deere & Co., for example, has been moving beyond equipment toward a digital platform that helps farmers make faster, better decisions.

“Instead of simply shipping equipment, they’re looking to own more of the lifecycle from predictive maintenance to data-driven optimization,” he said. “In many cases, that requires ecosystem collaborations and new capabilities that didn’t sit at the core of the organization five or 10 years ago.”

PwC survey respondents are expecting “fully 44%” of their companies’ total revenue to come from outside of manufacturing industrial or consumer products by 2030. The top three “growth hotspots” are technology, digital and communications products and services; defense, governmental and educational offerings; and energy and fuel production and distribution.

Fostering a strong company culture is also important for manufacturers on their adoption journeys. When leaders are confident about their digital transformation, but their frontline teams do not feel safe or supported in learning new skills, adoption slows, Hawk said.

To bridge the gap, manufacturers need to communicate effectively on how roles will change, invest in upskilling and create environments that encourage experimentation, he said. “Companies that treat workforce readiness as a core pillar of their automation strategy will be far better positioned to turn technology ambition into measurable performance gains.”

In a recently renovated office building in Watertown, Massachusetts, 100 robots are slowly learning how to pick up random objects, including everything from hand lotion bottles to packages of Welch’s Fruit Snacks.

Each human-like machine is equipped with four cameras — one at the head, at the chest and on both claws — and fixed to a stationary box. A few employees carrying smart devices inspect the machines’ work to see if they complete their tasks correctly.

At this point, most of them do not.

The robots, named Sonny after the “I, Robot” movie character, have only been operational for a few days, and it’s going to take several months of research and development before Tutor Intelligence has collected enough data for industrial applications, CEO Josh Gruenstein said at a facility tour April 22. Some day Sonny could be deployed at a Fortune 500 company’s operations.

Tutor has just moved to its new headquarters at Watertown, Massachusetts, located at the former home of Boston Scientific, along the Charles River. The historic site was previously a cotton mill in the 1800s.

“This is really an instrument of discovery for us to determine what are the best robot learning methods, and to start to build that capability set where we might be comfortable going to a customer,” he said about the robot fleet.

DF1 and the unique way Tutor teaches its robots

Data Factory 1, or DF1, is what Tutor is calling “the largest data factory in the United States.” The fleet of robots are learning autonomously how to move and manipulate objects with their claws, powered by the startup’s first vision-language-action model, Ti0, and through large-scale human supervision. In addition to the onsite team, remote workers in the U.S., Mexico and the Philippines monitor and correct the robot behaviors.

Tutor, founded in 2021, was started by Gruenstein and Alon Kosowsky-Sachs, chief technology officer, out of the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Laboratory. As students, they developed keen interests in the robot brain, and later set out to discover a “scalable recipe” for teaching robots how to do general tasks through their company.

The startup has raised a total of $42 million to scale its AI-powered robot workers, including a series A funding round of $34 million that closed in December.

Instead of using the latest wearable or sensor technologies, Tutor chose its recipe — a mix of cameras, vision-learning software and human supervisors — because it was more cost-effective.

“Actuators are expensive, sensors are expensive, even just like the physical structure of the robots … are not that big and could be 10 times cheaper,” Kosowsky-Sachs said. Cameras are cheap, in part, because of the iPhone and its massive scale of production, he said. He and Gruenstein hope that economies of scale will eventually take hold for other parts used in their robots.

“There’s this fundamental problem with robotics, which is, you can build the smartest robot and the world’s most capable robot, but if its [return on investment] isn’t extremely competitive, your business is uninvestable and will probably go out of business,” Kosowsky-Sachs said.

Support from AWS and its cloud services

In 2025, Tutor participated in the inaugural cohort of the Physical AI Fellowship, an eight-week startup accelerator supported by Amazon Web Services, Nvidia and MassRobotics, a nonprofit robotics hub located in Boston’s Seaport district.

As a member, Tutor received dedicated science and engineering support from AWS and Nvidia, as well as $200,000 in AWS credits to help “get their solution off the ground and then, of course, go to market,” said Alla Simoneau, physical AI technology lead at the AWS Generative AI Innovation Center.

Tutor’s approach to advanced robot learning differentiated it from other startups in the robotics and physical AI space, Simoneau said.

“What they were trying to do was build a brain for robots that could adapt to different environments,” she said. “And that generality was something that was quite interesting.”

The main goal of the fellowship, which has expanded globally this year, is to advance leading technologies. However, some of the participants have matured and scaled to the point where they can sell their services to customers through the AWS Partner Network, Simoneau said. Deployment in Amazon warehouses is not the end goal, she said, but it could be a byproduct.

“It’s critical to have a cloud partner that can support massive amounts of compute,” Gruenstein said. All of Tutor’s robots are connected to AWS.

Gruenstein said he thinks of Tutor as two halves: One is focused on developing fundamental robotics research and seeking how to build robot models and software, while the other is focused on robot deployments in factory and warehouse settings.

“I don’t think there are any other companies that are building the robot brain and then actually deploying it into the real world and doing productive work in that manufacturing context or that logistics context,” he said.

Tutor deploys its Cassie robot at US manufacturing and logistics facilities

While the Sonny robot is still in development, Tutor has deployed its Cassie robot for autonomous case-picking and palletizing with U.S. customers.

The 2,000-pound industrial robot can manipulate boxed goods and materials, among other tasks, for a wide range of customers, Gruenstein said. He declined to offer specifics about the cost and number of Cassie deployments. Early adopters in the food manufacturing and logistics sectors attended the tour and said they were already seeing the benefits from their pilot experiments.

Paul Baker, CFO at Productiv, a third-party logistics provider, said Tutor’s robot is operating at or better than the level of a human at its Dallas warehouse. The company is testing about 15 other robots that perform at sub-human levels, doing high-variability, high-dexterity tasks, with the goal of improving speed and efficiency.

“We’re working with groups, including Tutor, to try to [improve dexterity] as fast as possible,” Baker said. “That’s kind of that next frontier, and that’s why they have the data center working.”

Jeff Pulley, a facility manager at BetterBody Foods — a maker of peanut butter powder, avocado oil, chia seeds and other health-focused goods — said the company in July added two of Tutor’s palletizing robots to its Lindon, Utah, facility and one to a factory in Greenfield, Massachusetts.

Since implementing the robots, Pulley said BetterBody Foods has saved 36% compared to the cost of a worker that would normally be loading and unloading boxes from the pallets. He said the company hopes to add three more Cassie robots to its facilities over the next year and a half, and eventually add Sonny robots when they are ready.

“Robotics and automation, I believe, is … where everything’s going. If you want to delay it, you’re slowing down your own progression,” Pulley said. “So [you] might as well embrace it and learn from it.”

Sponsored content

By Rittal

Doing things better, much better, is what motivates Engineering Specialists. The systems integrator based in Brookfield Wisconsin, certainly doesn’t do things by halves. Its vision is to establish a holistic production system for the company’s own panel-building operations. The powerful impact of Eplan software and Rittal automation technology is already evident in the shape of enhanced productivity, higher quality and more attractive working conditions. 

An initial peek inside the production building, which covers an area of about 645,000+ sq. ft., already suggests that a lot of things here are different from what you would normally expect to see at an industrial production plant. It is bright, clean and quiet, ideal working conditions for the 140+ employees. Something else quickly becomes clear, too. Besides providing plenty of space for many, many enclosures, the company also has very big ideas and is looking a long way into the future. As demonstrated by its use of the latest data-driven automation systems, Engineering Specialists is run by innovators and visionaries who are determined to get the most out of their panel building operations right now – and are always thinking ahead. “I’m very much a visionary and see opportunities for doing things far better. When it comes to our customers’ automation processes, for example, we are helping to improve efficiency,” says John Miller, Vice President of Engineering Specialists. “When I took a look at our own panel building operations some years ago, however, I realized we weren’t doing a good job ourselves,” he adds. This realization set the transformation in motion.  

Precision and speed combined  

Although the first CNC system for machining large enclosures put an end to laborious and noisy manual processing using jigsaws, requirements increased further still. When the company moved into its new production building three years ago and the number of customers grew in a very short space of time, this suddenly created capacity problems.  

“We were machining enclosures on multiple sides and handling higher volumes and we just couldn’t keep up. Our lead time was simply too long,” reveals Miller. To resolve the bottleneck, the company purchased a Rittal Perforex Laser Centre LC.  

For the first time ever, it was now possible to machine enclosures quickly and from five sides in a single operation. This was particularly beneficial in the case of stainless steel enclosures, which the company supplies to numerous customers in the food and beverages sector in Wisconsin. “The LC is much faster than a vertical CNC machine and we don’t have all those residues and burrs to remove afterwards,” says Miller. What’s more, lots of projects require very precise, square cut-outs in the enclosures to insert small connectors. A task that previously involved manual drilling followed by laborious filing is now completed by the laser with maximum speed and precision. “The laser can machine an entire enclosure – from several sides – in just ten minutes,” he continues.  

Consistent wiring  

Miller then realized he had a further problem – the wiring process. “I asked myself how we could make it consistent. We build a control panel for one particular OEM every month, but no two employees wire in the same way. I wanted to achieve a professional result,” he explains. Miller was also looking to reduce the huge amount of wiring waste and speed things up by using pre-assembled and labelled wires. During a visit to the Rittal Application Center in Houston, he discovered the Rittal Wire Terminal WTC, a wire processing machine that has now been in operation at Engineering Specialists since 2024. 

Switch to Eplan  

That also involved a new approach on the software side. According to Miller, automatic wire processing called for a hard switch to Eplan at the engineering stage. Younger engineers at the company and a growing number of customers were, he says, already recognizing the benefits of Eplan over AutoCAD, which has been in use for over 40 years. 

“We can see the engineering efficiency achieved with Eplan – thanks to the Data Portal, for instance, but also the use of Eplan in the machine control system,” emphasizes Miller. “We’re starting more and more projects with Eplan and we want to use the software throughout the organization,” he adds. Miller describes his ultimate goal as follows: “Using the Eplan Platform to combine all the automation solutions in a first-rate, holistic production system that will maximize process efficiency.”  

In the future, Engineering Specialists also intends to use Eplan Smart Wiring and integrate it into the workflow. “The software helps staff get the app-based wiring process exactly right. Moving forward, that means we won't necessarily need to have specialist staff who can read wiring plans working in panel building operations,” says a delighted Miller. 

Nvidia on March 16 introduced its latest models and frameworks designed to help the world’s leading robotics companies scale physical artificial intelligence beyond experimentation to real industrial applications at its 2026 GTC conference in San Jose, California.

The technology giant is rolling out its new Isaac simulation frameworks, as well as its Cosmos and Gr00t open models for the industry to develop, train and deploy the next generation of intelligent robotics.

Industrial leaders such as ABB Robotics, Fanuc and Yaskawa are integrating Nvidia Omniverse libraries and Isaac simulation frameworks into their systems to validate robots and production lines using digital twins.

Nvidia, originally known for its gaming graphics cards, has become the poster child of the AI boom in recent years. In October 2025, the technology giant’s market valuation surpassed $5 trillion, Reuters reported. Its success has been driven by surging demand for premium graphics processing units that power data centers used largely for storage and AI factories, which are built specifically to run AI models.

 

In addition to chips, the company is also expanding its influence in the physical AI space as manufacturers look to make their workflows and production lines more efficient through automation and robotics. As industrial robotics become more AI-driven, manufacturers are looking for software that can accurately emulate what will happen on the shop floor before deployment, according to Nvidia. Developers are leveraging Nvidia’s Omniverse libraries and Cosmos to create more physically accurate simulations, as well as Gr00t to give robots a “human-like” brain.

 

“NVIDIA’s full-stack platform — spanning computing, open models and software frameworks — is the foundation for the robotics industry, uniting a worldwide ecosystem to build the intelligent machines that will power the next generation of factories, logistics, transportation and infrastructure,” CEO and founder Jensen Huang said in a statement.

 

Some of the leading robotics companies are using Nvidia’s software to support enhanced simulation, digital twin and sensor processing capabilities, including European-based Universal Robots and Kuka.

 

“We’re working with them to implement our physical AI models [and] integrate into simulation systems so that we could deploy these robots into manufacturing lines all over,” Huang said during his keynote presentation at GTC.

 

U.S. developers Field AI and Skild AI are both using Nvidia’s models and frameworks to build robot brains for accelerated training. Pittsburgh-based Skild AI is working with ABB Robotics and Universal Robots to deploy its robot intelligence across different industries, according to a news release. It is also partnering with contract manufacturer Foxconn to enhance its production of electronic devices such as iPhones and Nintendo consoles.

 

Workr, marketed as a “robotic workforce company,” is leveraging Nvidia’s Omniverse libraries to train robots that can be deployed by small- and medium-sized manufacturers in minutes without programming knowledge. It is also integrating its AI platform ABB Robotics.

 

Other technology companies, such as Microsoft and Google, are leveraging Nvidia technology to develop their own models and lay the groundwork for physical AI applications. Manufacturers are projected to more than double their use of AI and automation by 2030, according to a recent PwC survey. One of the biggest hurdles for them to overcome is getting their workforce trained and ready for the transition.

 

Eventually, Huang said, “every industrial company will become a robotics company.”

 

As manufacturers increasingly invest in digitization, automation and artificial intelligence, they’re also learning how to manage an exponential increase in data.

Speakers at multiple panels during a May 6 event hosted by the Massachusetts Institute of Technology’s Initiative for New Manufacturing discussed how they’re thinking about data standardization and analytics. They’re also trying to be strategic about how best to approach this evolution in technical capabilities.

“Automation, AI, analytics ... by themselves, do not create value. I believe they actually increase the flexibility or the domain of what's possible, and they'll only create value when they're applied to a well-defined problem,” said Swamy Kotagiri, CEO of automotive supplier Magna International. “Otherwise, I like to say you're experimenting and you're not transforming."

Automaker Ford is also investing in automation and digitization to gain a competitive advantage. However, Yung Fung, Ford’s managing director and general manager of advanced industrial technology and platforms, cautioned that more data isn’t always useful without the right approach.

"We have a bunch of things that we're trying to land to drive productivity — whether it be robots, AI systems — and everything produces a signal. There's a lot of signal in the factory. Too much data actually in a lot of cases,” said Fung. “The question is are there ways where we could have more connectivity between all those signals.” 

Ford and other manufacturers described throughout the event how managing these disparate sources of data is an ongoing task. Carpenter Technology, a global alloys manufacturer, hired its first chief digital officer in 2021 with this in mind.

“Data has always been a byproduct of manufacturing, and that's how it has been treated. We need to start treating data as an asset,” said Chief Digital Officer Shakthi Logasundaram, citing data as “the number one challenge” for the company’s technical operations.

Logasundaram said this goes beyond focusing on enterprise resource planning systems and financial data. He said the company is also tackling the harder task of organizing data from disparate machines and manufacturing processes for which “there's no standard, there's no control” because  “every machine is unique, everything is special.”

He said Carpenter chose to pull back on certain technology spending to first focus on standardizing data across the shop floor and fully organizing its system. 

“We are taking a completely different approach of scaling up and building an end-to-end data model of our entire organization. And [the] data model is not just going to come from the data that we have collected. We have 240,000 pages of critical process, variable data in paper that we have to digitize,” he said, citing the potential for generative AI to help with this task and agentic AI to help disperse the knowledge.

Biopharma company Amgen has also taken on the complex task of standardizing its data.

“I run into instances where I think I have the right data, and then I go dig into it a little bit more only to find, well, that's not completely the data. There's another set of data over here,” said Arleen Paulino, SVP of global manufacturing. “If you really wanted to leverage the technologies that are in front of us and are coming with AI, your data has to be really accessible and visible.”

Paulino and others also talked about how there can be initial challenges when certain divisions within a company may want to use unique organizational systems for their data, or may be hesitant to share it in general.

Sanjay Sharma, CEO of ArcelorMittal China, spoke about how his company is aiming to “democratize the data” while also navigating regulatory factors.

“Everybody wants to hold the data, and geopolitics makes it even more difficult — transportability of data from one region to another region, that brings certain barriers also,” he said.

This effort also requires an investment in workforce transformation to hire and train data scientists or related positions, Sharma added.

While the required investments and culture changes take time, manufacturers also said they see clear benefits to this digital shift.

Bill Good, VP of supply chains at GE Appliances, said he views this as one way to compete against countries with lower-cost labor, alongside efficient domestic supply chains.

Good said the company has largely retrofitted its older plants to have full data tracking, demonstrating how he can track production metrics at company plants from an app on his phone.

“I literally have visibility to every single assembly line, every piece of equipment in my plants,” he said. “Honestly, for me, I don't need visibility into everything, but my folks do. I often tell people the most difficult problem is the problem you can't see, and data allows you to eliminate the problem. So we have focused tremendously on that digitization strategy.”

Paulino also discussed how Amgen takes a “control tower” approach by delegating data responsibility to different tiers of people. Those on the plant floor focus on information about their respective areas and a plant manager can have sitewide visibility. Certain executives can also look at trends more broadly, while keeping in mind how and when to act on what they’re seeing.

“There's a responsibility also that comes with having access to all of that data. How do you use it in a way that's beneficial and not a hindrance? Because nobody wants me calling them all the time, saying, ‘Hey, I'm watching your run and it looks like it's going off track.’”

As all of this evolves, companies are figuring out how best to manage proprietary data versus data shared with third-party technology providers.

“What we want to keep in house is anything that there is an expertise that we have ... that part of AI we want to keep inside,” said Logasundaram, versus other use cases that are more suited for third-party technology providers. For example, he said when it comes to “predicting machine failures, anyone can do that.”

While data is often thought of as a way to track the performance of certain equipment, it can also take many other forms.

"The secret sauce for any plant ... is the conversations that go to problem solve and understand and triangulate the context. That goes into the ether,” said Ford’s Fung, adding that “it's not captured in a database, it's not captured in a report" because the people that are working don't have time to do it.

"How do you start to capture that which is more powerful than sometimes a raw dataset because it has context to create your semantic anthology?” he asked. "That's something we're also exploring and working on. Because we have scale, we have a ton of data. We're not harnessing all that context.”

For years, the term collaborative robot — or cobot — referred to a robotic arm or machine that could work alongside humans. But not anymore.

Recent safety standard revisions, ISO 10218 and ANSI/A3 R15.06, move away from labeling a machine itself as a “cobot” and instead require safety to be defined at the application level by using the term “collaborative application.”

This is the first major update to standards over the past decade and will significantly change how these tools are designed, certified and marketed, as well as how manufacturers implement them in collaborative human-machine processes with more strictly regulated safety measures.

John Jakomin, chief engineer and co-founder of Slip Robotics, a company offering robots-as-a-service to manufacturers, called the change an “overdue correction.” 

“Collaboration is a feature of how the robot is utilized, not a property of the individual robot itself,” he said. “The revised wording acknowledges that risk and safety are based upon the entire system, including the robot, tools, workplace, layout, control systems, operator's proximity and other complementing engineering controls.” 

In other words, it’s no longer sufficient to buy a cobot off the shelf and assume safe human coworking conditions. Instead, an application-specific risk assessment has to be carried out and documentation of how humans and robots will interact becomes central.

The new guidelines divide the safety and capabilities of collaborative applications into four categories: power and force limiting, speed‑and‑separation monitoring, safety-rated monitored stop and hand guiding.

“If a robot is using a sharp tool, it’s not truly 'collaborative' unless safeguards are in place,” said Jasmeet Singh, executive vice president and global head of manufacturing at Infosys. “To keep people safe, you need measures like monitored stops, power and force limits, emergency stop buttons and speed reduction when someone is nearby.”

To support this, Infosys has developed the Integrated Robotic Assisted Test Automation Framework, specifically designed for bots to automate testing and validation of human-machine interface devices.

Cobot technology has come a long way in recent years, Singh said. “Thanks to advances in vision systems, sensors and perception technologies, these robots now have real-time situational awareness and can interact with humans more naturally,” he added. “Add 3D vision, lidar and smarter control algorithms, and you’ve got systems that are highly responsive and aware of their surroundings.”

Jakomin noted that system design is more important for safety than just sensors or software.

“This means making sure that operators comprehend robot behavior, establish safe zones, monitor with light curtains or vision systems and develop recovery procedures for unforeseen circumstances,” he said.

The real game changer here, according to Singh, is AI and machine learning.

“Cobots can now make intelligent decisions, recognize objects dynamically, plan movements adaptively, predict behaviors, and collaborate with humans in real time,” he said.

Freddy Kuo, chairman at Luminys and special office executive assistant at Foxlink is among the people helping develop reasoning-based artificial intelligence models, or the “brains” that enable robots to make context-aware decisions and adapt in real time. 

According to him, while robot makers and the manufacturers that implement them revise systems to match the new safety standards, there’s another easy-to-dismiss but critical angle to consider: cybersecurity. 

“Physical safety — like sensors, force feedback and automated shutdown systems — has long been a defining feature of collaborative robots, but now with cobots starting to use cloud-based AI systems, we must think carefully about data security, intellectual property protection and data integrity,” Kuo said. “Every time a cobot sends or receives data for model updates or training, it introduces potential vulnerabilities. So now manufacturers must treat safety as a two-part responsibility: protecting both people and information.”

While manufacturers across the U.S. are at different stages of cobot adoption, ongoing labor challenges  and economic uncertainty are motivating more companies to automate tasks, especially as the cost of deployment goes down, Jakomin said. 

IMARC forecasts the U.S. cobot market to reach over $8 billion by 2033, with a compound annual growth rate of approximately 28% from 2025 to 2033. However, the U.S. industry is still relatively small compared to global robot deployment, as many small-to-mid-sized manufacturers have yet to implement cobots and related tools.

Globally, the cobot market has already crossed $3.5 billion with nearly 150,000 units shipped in 2025, according to Omdia, which expects the market to more than double in just the next three years.

Right now, Kuo said, large-scale facilities with steady, predictable processes are the easiest environments to introduce cobots, but the real power of this technology lies in adaptability.

“Cobots will be able to learn and adjust to virtually any environment or task. That means they won’t just belong in high-volume factories, they’ll become valuable in smaller, more specialized settings, too,” he said. 

As the technology becomes more intelligent and accessible over the years, Kuo predicts that “smaller companies will soon gain the ability to build, customize and produce independently, leading to a much more decentralized and flexible manufacturing industry.”

While AI and automation seem to be the biggest trends in the industry, Intrinsic Chief Technology Officer Brian Gerkey recently shared a striking statistic: 80% of U.S. manufacturing facilities have zero automation.

In spite of discussion about the potential benefits of this technology, it’s still far from widespread in the United States, let alone at the level of fully automated factories as seen in countries like China and Japan.

“There is no doubt interest is high across the board, but execution is where things get difficult,” said Jeff Burnstein, president of the Association for Advancing Automation. The group’s research also shows that while a strong majority of manufacturers believe AI will be critical to their future, only a small percentage say it’s widely deployed today.

Deloitte’s 2025 Smart Manufacturing and Operations Survey showed similar results. An estimated 92% of manufacturers surveyed said they believed smart manufacturing will be the main driver for competitiveness over the next three years.

Yet only about 29% of manufacturers reported already using AI or machine learning at the facility or network level, and only 24% had deployed generative AI. Looking ahead over the next two years, 41% of respondents said they planned to prioritize factory automation investments.

The bottlenecks holding back AI adoption

Manufacturers are still building the foundational capabilities required to scale AI and automation, said Tim Gaus, a principal and smart manufacturing leader at Deloitte.

Despite nearly three in four companies planning to deploy agentic AI within two years, only one in five reported having an equipped model, according to the company’s State of AI in the Enterprise report.

“Many organizations are still working with fragmented legacy systems and data that is not structured for AI use,” said Jasmeet Singh, executive vice president and global head of manufacturing at Infosys.

According to Singh, it often comes down to digital maturity. Manufacturers that have already modernized their core systems, invested in cloud and built strong data foundations are moving faster on AI. Singh said such companies are better positioned to scale beyond pilots because their data is ready to support advanced use cases.

Singh added that a key shift where many manufacturers struggle is the transition from pilot projects to measurable business outcomes, because they want a clear return on investment before committing significant funds.

“In many cases, earlier proof-of-concept efforts did not translate into enterprise-wide impact, which has slowed broader adoption,” he said.

The disappointment often comes from how AI was implemented rather than the technology itself, something Singh called using “AI for AI’s sake.”

But even these “failed” projects can be valuable, according to Burnstein, because they give “a much clearer understanding of where the real constraints are, which leads to more successful implementations the next time.”

Automation considerations

Most automation today is built for highly standardized, repeatable processes, said Stefan Nusser, chief product and commercial officer at Intrinsic, Google's robotics software company. But manufacturing often involves variability, customization and evolving processes, which can make automation expensive to test and implement.

Even when they’re established, Nusser said the processes may run for only a short period despite the significant bandwidth required to set them up. This may be because automation is just used for a specific task but not scaled across the whole factory. Companies may also realize the technology is not feasible or profitable for their current business model, or their manufacturing needs change over time and require adaptation to newer processes.

Major bottlenecks are commonly seen with mid-sized manufacturers that have the resources to experiment with AI and automation solutions but tend to wait until the value is clear before making large-scale investments. On the other hand, larger organizations have more resources to pursue aggressive growth, but their scale and legacy systems can slow things down, Singh said. 

That’s why Nusser advocates for incremental adoption. He cited the example of autonomous mobile robots that are flexible and less risky to implement, compared to a legacy, conveyor-based solution that requires overhauling the entire structure and workflow. The latter may seem nearly impossible, according to experts, considering that shutting down the system for even a few days could cause losses and supply chain disruptions in the millions. 

History repeats itself

This is not the first time the manufacturing industry has experienced a complicated technology transition. It also happened with the adoption of simpler robots.

“The technology faced resistance because of initial high costs and hesitancy around change management, but eventually, sectors like automotive manufacturing saw huge benefits after prioritizing the foundations needed to make it effective,” Gaus said. 

AI and advanced automation are following a similar trajectory. Early adopters are demonstrating clear productivity gains, he said, with some companies reporting 10% to 20% improvements in production output and employee productivity.

“As these results become more visible, we can expect adoption to accelerate across the industry.”

According to a recent PwC report, manufacturers could triple their current automation levels by 2030. Some use cases showing the strongest potential for adoption of automation in the next few years include data capture and analysis, demand forecasting and quality control.

A recent report from the International Federation of Robotics shows that trend has already been building in recent years. An estimated 542,000 industrial robots were installed across the world in 2024, more than double the number 10 years prior. U.S. installations reached a peak of 44,000 units in 2023.

Their findings echo Gaus’ predictions: “One thing we consistently see is that once manufacturers observe successful use cases — especially from their peers — adoption tends to accelerate very quickly.”

While the increasing use of AI has been making news in the U.S. manufacturing industry, countries such as China and Japan have been running fully automated factories that require minimal human intervention for decades.

These so-called “dark” or “lights-out” factories use machines, robots and automated systems to operate 24/7 without breaks, lighting, heating or other conditions humans require.

A famous example is a Fanuc factory in Japan, which reportedly runs for a month without much human input to build robots at the rate of 50 per 24-hour shift. Philips is doing something similar at its razor-manufacturing plant in the Netherlands, where 120-plus robots run the fully automated factory and require fewer than 10 workers for quality assurance. 

Could the U.S. implement such concepts to help alleviate some of the industry’s persistent issues, like labor shortages and supply chain disruptions?

Increasing use of AI and automation in the U.S.

U.S. manufacturing companies have increasingly begun experimenting with automation in recent years as the pandemic drove interest in artificial intelligence and digitization. However, adoption is slower and there’s no fully automated plant in the region as of yet. 

According to experts, the closest U.S. manufacturers have come to “dark” factories is Tesla’s heavily automated gigafactories. The company reports 90% automation at its Nevada plant, according to the Reno Gazette Journal, and even higher in China.

“Most U.S. manufacturers pursue gradual automation — they extend lights-out operations step by step rather than build entirely new dark factories,” said Daryl Edwards, a former plant manager and the founder of Agent Impact, a company working with manufacturers to implement AI solutions.

This is the opposite approach to China and Japan, where many lights-out factories were built from the ground up for new products, said Craig Melrose, global managing partner of advanced technologies and mobility at Htec, a company providing AI and embedded product engineering services.

“In contrast, the U.S. would largely be retrofitting existing facilities, which were designed for human operations,” said Melrose.

This comes with its own upfront costs, which many companies may not be ready to bear. 

“Implementing automation is expensive and most manufacturers must phase it in gradually to make it financially viable and ensure a clear return on investment,” said Stephanie Scearce, vice president of workforce innovation at the Georgia Association of Manufacturers, via email.

“Budget constraints, organizational understanding, process continuity, and (perhaps most importantly) leadership buy-in all play critical roles in adoption,” she said. “Aligning these factors is challenging and achieving that often takes significant time and resources.”

That’s why hybrid models are on the rise in the U.S., where some aspects of manufacturing are automated under human supervision.

“We’re not going to see a whole shift to dark factories in the next five years, but hybrid models with smart automation, digital twins, and AI copilots are accelerating rapidly,” said Alex Shikany, executive vice president of the Association for Advancing Automation, or A3.

That has given rise to collaborative robots, known as cobots, which require a tech-trained human operator. In Q2 , nearly a quarter of the units ordered in North America were “collaborative,” Shikany said, noting larger demand from automotive manufacturers. “They’re a clear signal that manufacturers are building automation with people, not just instead of them.”

Edwards interviewed 20 manufacturing executives about their automation journeys and found that one of his clients, a discrete manufacturing company, had begun implementing lights-out manufacturing at night.

“In the last hour or two of every shift, they initiate operations in all the semi-automatic machinery like CNC machines so they run lights-out overnight unsupervised,” he said.

This benefits the operators who prefer daytime shifts and want to avoid shift rotation. It also benefits the company since it can manufacture much more in a given time period without increasing costs, Edwards added. 

Currently, automation is the most common application in U.S. manufacturing sectors with high-volume, repeatable tasks such as electronics, automotive components, semiconductors and pharmaceuticals.

These sectors benefit from precision, consistency and around-the-clock uptime, Shikany said. As AI-powered digital twins improve, he added, dark-factory readiness will grow in industries where safety and optimization are critical, like food and beverage, and chemical manufacturing. 

Is AI taking away jobs?

When asked about what happens to the workers as factories increasingly adopt automation, experts said it doesn't eliminate the need for humans, but shifts their roles.

“Even in a highly automated environment, people are still essential for maintaining equipment, programming systems, troubleshooting issues and managing the complex touchpoints that make manufacturing run smoothly,” Scearce said. “Rather than aiming for completely unmanned facilities, I see the real opportunity in leveraging automation to create a balance between our shrinking labor supply and the growing demands of manufacturing.”

In practice, Scearce said, “automation will likely replace the types of manufacturing jobs we struggle to retain, which are often entry-level, highly repetitive, or physically demanding roles and jobs that fall into the ‘4 Ds’: Dark, dirty, dull and dangerous.”

Changes in the workforce are inevitable as more companies adopt AI and automation, but that doesn’t necessarily have to mean sending all the workers home. 

This is apparent in the upcoming U.S. factories as companies are hiring workers to implement certain automation systems as well as continuing manual processes still untouched by AI. 

“Our industry’s focus should be on upskilling and reskilling, not replacement,” Shikany said. “The manufacturing jobs of the future involve operating and maintaining automation systems, interpreting data from AI models and designing new workflows.”

To help with this change, A3 has begun offering certification and training programs in robotics, machine vision, motion control, and automation. The organization also hosts conferences and events, webinars, and online courses focused on implementation solutions. 

Many community and technical colleges, as well as company training centers, have started similar initiatives to match the growing need for an AI-ready workforce. 

Such transformation is natural and unavoidable as industries evolve from hand tools to mass production, and now to AI and lights-out manufacturing, Melrose said. 

“Digital was considered the four industrial revolution and AI could be viewed as the fifth,” he added. “With each revolution, workers were asked to change, adapt and learn new skills, which will continue to be true.”