Artificial Intelligence Clashes with Industry: The Gap Threatening Factories

In recent years, artificial intelligence (AI) has become a central field for technology companies and startups, but the transition from development labs to the factory floor poses significant challenges. According to Tsuri Dabus, chairman of the company Klil, the gap between the high-tech world and industrial reality could delay the implementation of technology in factories.

Dabus, who is familiar with both worlds, describes a significant gap between the development environment and the conditions in which AI solutions must actually operate.

"I wear two hats, a high-tech hat and an industrialist hat, and the connection between them is full of challenges," he says.

One of the central problems is the gap between technological demonstrations and the reality on the production line. "Someone shows a demo where everything works and looks cool, but when a solution reaches the production line, you discover a different world." Factories deal with noise, dust, humidity, changing temperatures, and the constant movement of workers, forklifts, and cranes.

Another issue is the lack of familiarity with the professional language and regulations practiced in the industry. Dabus believes that entrepreneurs should know in advance the standards and terms accepted in each industry, and can even use AI tools to learn them.

However, the most significant gap lies in the level of accuracy required. While in the high-tech world it is sometimes possible to launch a product in beta and improve it later, on the factory floor, errors can have significant consequences.

"80% accuracy is not enough for the industry," says Dabus, emphasizing that an industrial solution must reach accuracy levels approaching 100%.

An error in the system can lead to defective products, production line shutdowns, damage to equipment, and significant costs. Alongside this, there is the question of the costs of using AI models, including processing and token costs, as well as the need to regulate in advance the issue of ownership of the data fed into the systems.


Dabus suggests that entrepreneurs focus on a defined problem rather than promising a solution to all challenges, set clear success metrics in advance, start with a limited pilot that does not jeopardize core operations, and examine partnership models that reduce the initial risk for the factory.

Alongside the difficulties, Dabus emphasizes that the potential of AI in industry is significant, and the market continues to invest in ventures in the field. The responsibility does not end with the success of a single project.

"Every failure of an AI system on the factory floor burns the trust of industrialists and makes it harder for those who follow," he concludes.