"Everything must be AI." Today, AI has become the political correctness across industries, including automotive. The wave of electrification and intelligence is sweeping through the automotive sector, fueling an intense pursuit of high automation. Whether it's the high automation in production processes—dark factories—or the high automation in vehicle functions—smart cockpits and intelligent driving—it has become a focal point in the industry today.

The application of AI in the automotive industry can be divided into the consumer side and the production side. However, regardless of which side it is, AI faces practical challenges, sparking debates and skepticism from various stakeholders, and even harboring potential "pitfalls."  On the consumer side, the debate over intelligent assisted driving requires no further elaboration. While many tech-savvy early adopters advocate for it, a significant number of experienced drivers still claim they "never trust machines and will never hand over the steering wheel."  On the production side, AI-driven high-precision manufacturing and inspection are regarded as "capable of improving vehicle quality and production efficiency." Yet, in reality, companies that were among the first to deploy AI technologies in factories have not succeeded in reducing defects or the frequency of vehicle recalls.

It is still the US market that has pioneered the use of data experience to "explore" the world. Ford and Hyundai have become the first two car companies in the US to use AI in car manufacturing, with each AI factory costing billions of dollars. It can even be predicted that the investment related to AI manufacturing in the entire industry may reach the level of billions of yuan. But before the end of this year, Ford has initiated hundreds of recalls in the United States due to product malfunctions, involving millions of vehicles. Therefore, evaluating the true impact of artificial intelligence on the automotive manufacturing industry still requires more time and data accumulation. When car companies spend billions of dollars (equivalent to billions of yuan) to build AI driven "smart factories" and promise to improve quality and reduce recalls, the actual efficacy of this highly anticipated technology is far from clear. If the automation of automobile manufacturing was still centered around mechanical concepts in the previous stage, then this stage is more combined with AI concepts with a greater emphasis on software. The competition establishes a 'black light factory', and no country can compare with China. But outside of robotic arms, "electronic eyes" and "AI brains" are still the race tracks that overseas car factories can compete on.

Ford and Hyundai are regarded as the leaders in this round of AI manufacturing competition in the US automotive industry. They claim that AI systems can accurately capture defects before delivering vehicles to customers. The Metaplant plant plant in Georgia, built with an investment of $7.6 billion by Hyundai Motor, has produced the first batch of Ioniq 5 electric vehicles in October 2024. Hyundai has introduced AI based Boston Dynamics robots, automated handling vehicles, and digital twin technology here, and has expanded its cooperation with Nvidia, continuously increasing its focus on "smart factories". The AiTriz vision system, launched by Ford in December 2024, has been installed in 900 workstations across 35 US factories. This system can verify component installation in complex situations where the line of sight is obstructed, which is more advanced than the first generation mobile vision system at the beginning of the year.

Ford sees AI as one of the key potential solutions to break through the game. Jeff Tonabene, Technical Manager of its Manufacturing Technology and Development Center, bluntly stated, "Our goal is to design products that cannot be installed incorrectly or go through the production process incorrectly." According to Ford's statement, quality improvements have been observed in areas where AI vision systems have been deployed and are being expanded in application, but it is also cautious to point out that "not all manufacturing processes are suitable or require AI. Meanwhile, Modern spokesperson Miles Johnson outlined its vision: "AI has enormous potential in improving quality control and early defect detection. ”But a sharp reality is right in front of us: even if AI technology is deployed in some factories, Ford Motor Company will still initiate 134 recalls in 2025, involving over 12.17 million vehicles. For Ford, improving quality and reducing pressure through recalls are urgent. This veteran giant, headquartered in Dearborn, Michigan, not only set a record for annual recalls in the US market, but is also likely to "top" the annual recall list for the fourth time in five years. When car companies loudly proclaim that "AI helps improve the quality of automotive products," industry observers remain cautious. Analysts believe that it is too early to evaluate the returns on these massive investments. Based on the current performance, the effect of AI intervention in factories is not ideal. The most publicly available quality data - recall announcements - give complex signals.

In the vehicles produced by Ford after deploying AiTriz, the average recall scale involving components is much lower than the annual average. However, during the same period, Ford still initiated 22 manufacturing defect recalls, including issues such as improper bolt torque and poor windshield installation, which are precisely the defects that AI vision and torque monitoring systems aim to capture. Ford did not specify whether the recalled vehicles were produced by AI workstations. Its spokesperson cleverly likened measuring AI through recall data to "questioning why the front door camera did not detect the rear door intruder". Ultimately, Ford's record breaking 134 recalls in 2025 affected over 12.17 million vehicles, with warranty expenses reaching $450 million in the third quarter of this year alone. Of course, Ford gave a very 'face watering' explanation - which reflects a reinforced strategy of 'high caution' and 'quick detection and repair of problems'. In contrast, Hyundai has initiated 14 recalls this year, including two involving Ioniq 5 produced at its Georgia factory, with issues focused on headlight focusing and wiring harness installation. In principle, neither Ford nor Hyundai has provided specific data that can directly prove that AI improves vehicle reliability. Both car companies coincidentally positioned AI as a 'long-term quality strategy' rather than an immediate solution. Firstly, there is a "fuzzy zone" between the use of AI in manufacturing and the factors affecting quality. It has broad prospects, but it is not a panacea, "AutoPacific analyst Robbie DeGraaf pointed out the key. He believes that AI tools may be able to help identify problems before large-scale recalls occur, but their analysis and identification capabilities may never reach 100% accuracy. As of the end of October, the National Highway Traffic Safety Administration recorded 817 recalls, which is lower than the 963 recalls recorded in the same period last year, but the overall number is still significant.

Secondly, AI is still in the early stages of deployment in automotive factories, and there are difficulties in evaluation and data mysteries. Measuring the impact of AI on quality has become a complex project that requires precise data and time accumulation. Adam Bernard, a former senior executive and industry analyst at General Motors, pointed out that "artificial intelligence" is a broad concept, and how it is applied determines what data is needed. Even J.D. Power, which publishes authoritative quality reports every year, has stated that it currently does not track the application of AI in factories. Therefore, we can conclude that AI does help improve product quality in certain fields, but "relying solely on AI to improve quality is actually dragging down quality. At the same time as AI demonstrates its "advanced" capabilities, we must also be cautious and avoid "traps". Both automobiles and automobile manufacturing cannot do without people. Automobile manufacturers must ensure that, in addition to AI, they maintain a people-oriented and solid post production vehicle analysis and supervision system. ”This is the common conclusion of AutoPacific analyst Robbie DeGroff and other experts, which means that there is an undeniable "human factor" in both the use of consumer car products (such as intelligent assisted driving) and the manufacturing of production cars. In the third quarter of this year, at the invitation of Lexus, the author visited Lexus Miyada Factory in Miyagi City, Fukuoka County, Kyushu, and Tanhara Factory and testing ground in Tanhara Town, Omei County, Aichi County, Honshu. The content displayed was not characterized by the high degree of automation of "people are completely replaced". On the contrary, with the collaboration of precision equipment and manual labor, the well-being and care of personnel have become the main focus.

For example, on the second production line of the Miyata factory, robotic arms are everywhere and AGV cars run efficiently, but the workers' figures are still orderly. Taking the installation process of windshield glass as an example, workers first clean the area where adhesive is to be applied. Then, the robot applies the correct amount of adhesive in the correct position, and the workers work in pairs to install the car window glass onto the vehicle. The highest level "craftsman" worker at Lexus factory explained that "applying the correct amount of adhesive in the correct position" is a field that robots are very skilled in. The task of assembling car window glass onto vehicles on the production line, which requires feeling and skill, is a job that people with sensitivity and flexibility are good at. Even in the same job program, humans and machines can highly integrate and complement each other, which is the secret to achieving high quality. In addition, the spraying process has some complex shapes and narrow spaces, making it difficult for large robotic arms to achieve manual spraying levels. More importantly, machine training cannot be separated from manual operation. If there are operational errors in the machine, manual backup correction is also required. And the final inspection process is also a fusion of manual and mechanical labor. It is said that there are 1800 places that need to be confirmed before leaving the factory.

Experts from both the East and the West have reached an objective consensus here. Paul Laver, Vice President of Innovation and Applied Technology at the University of New Haven, reminds that while the industry is pursuing technology, it is facing severe talent challenges. The pace of progress in manufacturing expertise may be faster than the speed of employee training, "he said," which could even weaken the effectiveness of state-of-the-art AI systems. When implementing AI, the leadership role of people must not be ignored. We must focus on improving employee skills and ensure that appropriate technology is adopted in the right way. ”At present, the journey of AI in automobile manufacturing has just set sail. Billions of yuan worth of funds have been thrown into the water, but the waves are still uncertain. Whether the promise of "zero defects" can be truly fulfilled, time and data will be the only touchstone.