How AI-powered robots are changing American manufacturing

According to the International Federation of Robotics, about 393,700 industrial robots were operating in US factories in 2024, up 3% year on year, while annual installations fell 9% to 34,200 units [R11]. Installations were still about 30% higher than ten years earlier. The United States runs the third-largest operational robot stock in the world, behind China and Japan.

The comparison with China is stark. China installed roughly 295,000 industrial robots in 2024, almost nine times the US figure, and had around five times as many robots in operation [R11]. For American manufacturers, the headline is not that automation has stalled, but that adoption is concentrated and that large parts of the industrial base remain lightly automated.

The automotive industry remains the pillar of US demand. The IFR’s preliminary figures recorded a 10.7% rise in car-industry installations to 13,700 units in 2024, and it describes the US car industry as one of the most automated in the world, while noting that other areas of manufacturing lag [R12]. Most robots installed in the US are imported, because relatively few robot manufacturers produce domestically.

A traditional industrial robot repeats a precisely programmed path. It is fast and reliable when parts arrive in the same position every time, and it struggles when they do not. Much of the cost of automation has historically been the engineering needed to make the environment predictable enough for the robot: fixtures, guarding, feeders and careful programming.

AI changes part of that equation. The IFR identifies physical, analytic and generative AI as a leading robotics trend: analytic AI combined with vision systems to inspect parts and optimise robot performance, vision-language models that help people instruct robots, and “physical AI” trained in simulated environments before deployment [R12]. In practice this shows up as bin picking of mixed parts, visual quality inspection, adaptive welding and robots that can be reconfigured with less specialist programming.

It is worth being precise about the limits. Vision-guided picking and AI-based inspection are established, commercially deployed applications. General-purpose humanoid robots working across a plant are not; even the IFR frames current humanoid work around single-purpose tasks [R12]. A credible automation plan starts with a bounded task, not a vision of a lights-out factory.

The strongest driver is labour. The 2024 Deloitte and Manufacturing Institute talent study estimated that US manufacturing could need up to 3.8 million workers between 2024 and 2033, and that more than 1.9 million of those jobs could go unfilled if workforce challenges are not addressed [R13]. About 65% of manufacturers surveyed said attracting and retaining talent was their primary business challenge.

The second driver is investment in domestic capacity. The same study reports that construction spending on manufacturing facilities nearly tripled from June 2020 and reached a record US$225 billion in January 2024, as companies sought to de-risk supply chains and responded to federal incentives [R13]. New plants are a natural moment to design automation in from the start, rather than retrofitting it into a crowded floor.

These are projections and survey responses, not guarantees. A shortage forecast is a reason to examine automation, not proof that a specific robot cell will pay back in a specific plant.

The most cited US study is Acemoglu and Restrepo’s analysis in the Journal of Political Economy. Using local exposure to industrial robots across US commuting zones, it found that one more robot per thousand workers reduced the employment-to-population ratio by 0.2 percentage points and wages by 0.42% [R14]. The effects were concentrated in the most exposed local labour markets.

That finding deserves to be taken seriously, and also read carefully. The study covers robot adoption from 1990 to 2007, before most modern AI-enabled systems existed, and it measures local labour-market effects rather than the fortunes of individual firms. It does not establish that today’s AI-guided robots will have the same effects, larger ones or smaller ones.

The more recent talent study points to a shift in the skills manufacturers seek, including data scientists, software developers, engineers and industrial maintenance technicians [R13]. A reasonable reading of the evidence together is that automation can both displace some tasks and raise demand for others; who benefits depends heavily on training, location and how deliberately companies manage the transition.

Start with the task, not the technology. Which job is dangerous, repetitive, hard to staff or a quality bottleneck? How much does the part, the position or the product mix vary? AI helps most where variation previously made automation impractical, and helps least where a simple fixture would solve the problem.

Count the full cost. A robot arm is often a minority of the project cost once integration, tooling, safety, software, downtime during installation and ongoing maintenance are included. Ask who will own the system after the integrator leaves, and whether the plant has the maintenance and data skills to keep an AI vision model accurate as parts and lighting change.

Define success before the pilot. Agree measures such as throughput, first-pass yield, scrap, safety incidents, overtime and changeover time, with a baseline and an observation window. Involve the operators who do the work today; they usually know where the real variation and failure points are.

For sales professionals, AI robotics is changing who sits in the buying conversation. Decisions that once belonged to a plant engineer increasingly involve operations, finance, IT, safety and HR, because automation touches data, workforce planning and capital budgets at once. Discovery that only reaches one of those people is incomplete.

The most credible sellers in this market lead with the buyer’s operational problem and measurable baseline, are honest about integration effort and limits, and propose a bounded pilot instead of a transformation story. Claims about payback or headcount savings should be tied to the customer’s own numbers, not to an industry average.

AI-powered robots are making automation practical for tasks that used to be too variable for machines, and US manufacturers have strong reasons to invest: labour shortages, new plant construction and competitive pressure. The evidence also shows that adoption remains uneven and that the workforce effects are real and unevenly distributed.

The manufacturers most likely to benefit will treat automation as an operating decision, with a clear task, a full cost picture, a measured pilot and a plan for the people whose work changes, rather than as a technology purchase.