AI Robots Learn Japan’s Confectionery Skills as Food Industry Faces Labor Shortages

A Robot Learns the Art of Baumkuchen

In Kobe, a robot named Theo is being trained to make baumkuchen, the ring shaped German cake that has been part of Japan’s confectionery culture for more than a century. The machine is designed to copy the movements and decisions of skilled pastry workers, including how long each thin layer of batter should cook before the next one is added.

The project shows how artificial intelligence is moving beyond office software and factory assembly lines. In Japan, AI is being connected to cameras, image sensors, cooking equipment and food production records to preserve skills that were once passed from master to apprentice over many years.

Juchheim Co., the Kobe based confectionery company behind Theo, opened a training facility and showroom for the robot at its headquarters on March 25. The company displayed the system at the 2025 Osaka Kansai Expo and is now lending it to about 20 businesses across Japan. Juchheim aims to raise that number to 100 by the end of the fiscal year.

Baumkuchen was introduced to Japan in 1919 by Karl Juchheim, a German pastry chef who founded the company. Its preparation is unusually demanding. Batter is spread over a rotating spit, cooked in an oven, and layered again and again until the cake forms its distinctive tree ring pattern. Small changes in temperature, humidity and cooking time can affect the result, which is why workers may need several years of practice to master the process.

How Does Theo Copy a Confectioner’s Judgment?

Theo uses cameras and image sensors to observe the work of experienced Juchheim employees. The system records visible features such as the color, thickness and texture of each layer, then links those observations to conditions inside the oven and the movement of the rotating spit.

This information allows the robot to estimate when a layer is ready and adjust the process for changing conditions. A human confectioner may respond to the feel of the batter, the appearance of the surface or the atmosphere in the kitchen. Theo attempts to translate those judgments into data that a machine can analyze.

The technology relies on deep learning, a form of AI training modeled loosely on networks in the human brain. Instead of receiving a complete list of instructions for every possible situation, an AI system studies a large collection of examples and identifies patterns. Once trained, it can compare new images or sensor readings with those patterns and select a suitable response.

Juchheim says Theo can learn a new technique in a few days, a much shorter period than the time normally required for a person to gain the same experience. The machine is not simply repeating one fixed motion. Its value comes from adjusting the process when temperature, humidity or ingredients vary.

Why Is Japan Turning to Cooking Robots?

The food industry is facing a sharper labor shortage than many other parts of the economy. Data from Japan’s Health, Labor and Welfare Ministry showed a ratio of 2.31 job openings for every job seeker in food service in February. The average across all industries was 1.13.

A high ratio means employers are competing for a limited pool of workers. Restaurants, bakeries and food manufacturers must also deal with an aging workforce and fewer young people entering skilled occupations. In some businesses, the problem is not simply finding staff for today’s shifts. It is finding people who can take over when experienced workers retire.

Juchheim President Hideo Kawamoto said the robot could support an industry struggling to find successors. His statement reflects the central reason companies are testing AI in kitchens: automation can help maintain production when there are too few trained workers to fill every role.

Japan’s wider demographic decline gives these experiments a national dimension. The government has discussed deploying about 10 million AI equipped robots across 18 sectors by 2040. Plans for physical AI focus on machines that can interpret their surroundings and act in the real world, rather than systems that only generate text or images.

What Is Tacit Knowledge, and Can It Be Digitized?

Many cooking skills are known as tacit knowledge. This means people can perform a task correctly without being able to explain every decision in a formal instruction manual. A veteran cook may know that a pan needs to be turned faster when ingredients release extra moisture, or that a cake layer is ready from a subtle change in color.

Traditional apprenticeships pass on this knowledge through repeated observation and practice. In Japan’s culinary culture, the idea that skills should be learned by watching an expert has remained strong. That approach can produce excellent results, yet it becomes fragile when there are fewer apprentices and fewer senior workers available to teach them.

Professor Hitoshi Matsubara of Kyoto Tachibana University, who specializes in AI development, said earlier AI systems were poorly suited to this kind of knowledge. He said advances in the technology now make it possible to record and reproduce movements that were once difficult to describe.

In the pre deep learning era, AI could not teach tacit knowledge, but as technology has advanced, it has become possible to digitize skilled workers’ movements, which involve tacit knowledge.

The process does not mean that every part of craftsmanship has been captured. AI depends on the quality of its training data, the range of conditions it has observed and the accuracy of its sensors. A robot trained in one kitchen may need more instruction before it can work reliably in another location with different ovens, ingredients or production targets.

From Baumkuchen to Fried Rice and Yakisoba

Juchheim is part of a broader movement to apply AI to food preparation. Tokyo based TechMagic develops robots that cook stir fried dishes such as fried rice and yakisoba. In 2025, the company introduced an app that calculates pan rotation speeds and other settings based on the recipe selected by a user.

The robots use the system to learn cooking times and heat levels for individual ingredients. TechMagic says this has improved the consistency of its machines. Such work is less about replacing a complete chef than about controlling a series of physical actions that are difficult to perform with the same accuracy during every shift.

Other Japanese companies are using AI to preserve skills outside the kitchen. In 2019, Dentsu developed Tuna Scope, which analyzes smartphone photographs of tuna tail sections and assesses their quality. The system is intended to replicate mekiki, the expert eye used to judge tuna. Human specialists may spend a decade developing that ability.

Kura Sushi later served tuna assessed through the system as AI Tuna at its conveyor belt sushi restaurants. In Hiroshima, Otafuku Sauce created an AI search tool in 2023 for its archive of sauce formulas. The system was trained on more than 15,000 products and prototypes. Searching for a formula that once took up to two hours can now be completed in about five minutes.

Physical AI Connects Robots to the Real World

Theo’s work sits within a larger effort to give machines more awareness of their surroundings. Conventional industrial robots usually perform carefully defined motions in controlled settings. Physical AI adds cameras, sensors and AI models so a robot can respond to variations in objects, heat, timing and space.

Japanese robotics companies are working with Nvidia, Fujitsu, Fanuc, Yaskawa Electric and Kawasaki Heavy Industries on this area. The partnerships combine Japan’s experience in precision machinery with advanced computing and AI models. Nvidia has presented tools intended to help robots understand physical environments and adapt to changing tasks.

Local computing capacity may also allow companies to train AI systems using sensitive production data without sending all of that information overseas. That matters to food manufacturers, which may want to protect recipes, process records and customer information. Secure networks, reliable sensors and trained technicians will still be needed before these systems can become common.

Japan’s food machinery industry has been building this support structure for years. Companies such as Nikko supply automation systems for seafood, meat and agricultural processing. Nikko also operates the Hokkaido Robot Laboratory, where students and workers can receive robot training and companies can test whether automation is suitable for a particular production site.

These training centers address a problem that could limit automation: businesses may have machines but lack people who know how to install, supervise, repair and update them. Robotics can reduce the number of workers needed for a task, yet it creates demand for engineers, operators and maintenance specialists.

Why Food Automation Is Harder Than Factory Automation

Restaurants and confectionery kitchens present conditions that are less predictable than many industrial plants. Ingredients change from batch to batch, menus vary, spaces are often crowded and customers expect food to look carefully prepared. Temperature and humidity can also affect dough, sauces and cooking times.

Cost is another barrier. A company must pay for the robot, sensors, software, installation, maintenance and staff training. Small restaurants may not have enough space for large equipment or enough daily production to justify the expense. Older machinery can also be difficult to connect to modern AI systems.

Service culture adds another layer of difficulty in Japan. Hospitality often depends on personal attention, presentation and quick responses to unusual requests. Robots can perform repetitive cooking or delivery tasks, yet they have not reached the full standard expected of human service workers. A robot may be highly effective at a narrow task while remaining poorly suited to a busy kitchen with many simultaneous demands.

Examples from other service settings show how the technology is spreading gradually. At Narita airport, a cat like robot has been used in an unattended souvenir shop as part of a demonstration aimed at reducing staffing needs. Restaurants and cafes have also tested robots for food delivery, pancake preparation, sushi rolling and okonomiyaki cooking.

Will Robots Replace Skilled Confectioners?

The immediate role of systems such as Theo is more likely to be support than total replacement. A robot can repeat a process, monitor conditions and preserve a recipe, while human workers continue to prepare ingredients, inspect results, handle unusual situations and develop new products.

That division could help companies protect traditional food techniques while making skilled production less dependent on a small number of individuals. It may also allow experienced confectioners to spend more time teaching, testing recipes and supervising several machines rather than performing every repetitive step themselves.

There are concerns about who controls the digital record of a craft. Once a worker’s movements and decisions are converted into training data, the company may own a valuable production system built from personal experience. Workers and employers may need clear agreements about consent, credit, pay and the use of that data.

Customers may also judge machine made food by different standards. Consistency can be a benefit, especially for a product that must look and taste the same across many locations. Some customers, however, may value the human story behind a traditional confection. Companies will need to explain how AI is being used and where human judgment remains part of the process.

Japan’s Experiment Has Wider Meaning

Japan is using food automation as a practical response to demographic pressure. The same approach can be seen in logistics, agriculture, retail, healthcare and manufacturing. The goal is to keep essential services operating with a smaller workforce, rather than waiting for population trends to reverse.

The country’s strength in robotics gives it a useful base for this effort. Japan has long produced industrial machines with high levels of precision, while its manufacturers hold large amounts of operational knowledge. Physical AI seeks to connect that hardware and experience with systems that can learn from visual and sensor data.

Success will depend on results that businesses can measure. Robots must reduce labor pressure, maintain food quality, work safely beside people and justify their cost. They also need support from education, reliable data infrastructure and workers who can manage the technology.

Theo’s planned expansion from about 20 users to 100 will provide a practical test of whether AI learned confectionery skills can work beyond a company showroom. If the system performs well in different kitchens, it could become a model for preserving other forms of culinary knowledge that are at risk of disappearing.

Key Points

• Juchheim’s Theo robot uses cameras and sensors to learn how skilled workers make baumkuchen.
• Baumkuchen requires repeated layering and careful control of heat, time, temperature and humidity.
• Japan’s food service job opening ratio was 2.31 per job seeker in February, compared with 1.13 across all industries.
• TechMagic, Dentsu and Otafuku Sauce are applying AI to cooking, seafood grading and recipe searches.
• Japan plans to expand physical AI and aims to deploy about 10 million AI equipped robots across 18 sectors by 2040.
• Robots may preserve skilled techniques, while human workers remain needed for supervision, creativity, maintenance and customer service.