Conclusion First: The Cost Structure of ‘Hiring People’ is Starting to Change Fundamentally

Google DeepMind has announced “Gemini Robotics 2,” a technology that integrates and controls the entire body of a robot—from its feet to its fingertips—using AI. At the same time, a startup is working on a project where chefs wear cameras to capture their cooking movements, allowing robots to learn through imitation.

When these two pieces of news are placed side by side, what emerges is a future where the ‘cost of hiring one person’ collapses.

A chef earning 5 million yen a year. A line worker earning 4 million yen a year. What happens when this ‘cost of hiring one person’ becomes replaceable by robots and AI? Is it a threat or an opportunity for small and medium-sized enterprises?

What Has Changed with Gemini Robotics 2

Traditional industrial robots were “machines that repeat fixed movements accurately.” Welding robots, painting robots, picking robots—all specialized in specific tasks and required teaching (pre-programming movements) that cost hundreds of thousands to millions of yen. They would stop working with even a slight change in the environment.

What Gemini Robotics 2 has changed is the ‘cost of teaching.’

Specifically, by issuing instructions in natural language, the robot autonomously plans and executes movements. In a demonstration with Apptronik’s humanoid robot “Apollo 2,” a simple instruction like “place this box on the shelf” led the robot to estimate the box’s size and weight, adjust its grip, and walk to the shelf to complete the task.

The important structural change here is that ‘teaching becomes unnecessary.’

Previously, introducing a robot involved not only the purchase price but also teaching costs of 1 to 3 million yen. Moreover, re-teaching was required every time the product type changed. This was the main reason small and medium-sized manufacturers have been hesitant to adopt robots. Once AI can automatically generate movement plans, this barrier will significantly lower.

Filming a Chef’s Movements → Robots Cooking. What’s Impressive?

Another noteworthy piece of news is the project that captures chefs’ cooking movements with cameras and trains robots using that data.

The essence of this method lies in the dramatic reduction of the cost of ‘digitalizing tacit knowledge.’

Until now, trying to transfer the skills of seasoned professionals to robots required motion capture suits (costing hundreds of thousands of yen per set), specialized studio filming, and data processing—costing tens of thousands of yen per movement.

Now, this is becoming possible with a combination of commercially available cameras and AI. Chefs cook as they normally would, and AI analyzes the video to extract hand movements, force, and timing for the robot to learn. The filming cost is nearly zero, and data processing is done in the cloud.

A world where ‘digitalizing tacit knowledge’ drops from 3 million yen to under 50,000 yen is coming into view.

Will the Cost of ‘Hiring One Person for 5 Million Yen’ Really Collapse? — Cost Estimation

Let’s consider the actual numbers.

Current Labor Costs (for one on-site worker in small and medium-sized manufacturing and food service):

• Annual Salary: 3.5 million to 5 million yen
• Company-Borne Social Insurance: About 15% of annual salary (approximately 500,000 to 750,000 yen)
• Recruitment Costs: 300,000 to 1 million yen per year (job advertisements, interview labor)
• Training Costs: Approximately 500,000 yen per year including OJT
• Total: 4.8 million to 7.25 million yen per year

Moreover, turnover is an issue. The turnover rate in local small and medium-sized enterprises is high. If an employee leaves within three years, this investment returns to zero.

Estimated Costs for Introducing Robots + AI (Forecast for Around 2027):

• Robot Arm: 2 million to 5 million yen (price range for collaborative robots)
• AI Control Software (SaaS): 50,000 to 150,000 yen per month (600,000 to 1.8 million yen annually)
• Initial Setup: 500,000 to 1 million yen
• Maintenance: 300,000 to 500,000 yen per year
• First Year: 3.4 million to 8.3 million yen; Second Year Onwards: 900,000 to 2.3 million yen

The running costs from the second year onwards are 1 million to 2.3 million yen annually. Compared to labor costs of 4.8 million to 7.25 million yen, this calculates to a payback period of 2 to 3 years.

Of course, many technologies are still at the ‘demo level’ at this point. However, the direction of the cost curve is clear. The price of the robot itself is decreasing by 10 to 15% each year. The cost of AI inference has dropped to less than one-tenth in the past two years.

When Will Small and Medium-Sized Manufacturing See This Change?

To be honest, this is not something that can be implemented immediately.

However, the timeline is becoming clearer.

2025-2026: Large Corporations’ Pilot Testing Phase
Pilot implementations using Gemini Robotics technology will begin at automobile manufacturers and large food factories. Insights will be gained on ‘what can and cannot be done’ during this phase.

2027-2028: SaaS and Packaging Phase
Based on the results of large corporations’ demonstrations, robot manufacturers and system integrators will start offering ‘packages for small and medium-sized enterprises.’ Robot + AI services available on a subscription basis will emerge. This will be a realistic entry point for small and medium-sized enterprises.

2029-2030: Phase Where ‘There’s No Reason Not to Implement’
Prices will drop further, making it clear that ‘it’s cheaper than hiring one person.’ Companies that have not adopted this technology by then will be at a disadvantage in terms of cost competitiveness.

In other words, the real game for small and medium-sized manufacturing will begin in 3 to 5 years. However, there are things that need to be prepared now.

Three Things Small Businesses Should Do Now

1. Identify ‘Tasks That Are Dependent on Specific Individuals’

The greatest impact of robots + AI will be in situations where ‘tasks that can only be performed by specific individuals’ can be replicated. The discussion about transferring a chef’s cooking movements to a robot is a prime example of this.

What tasks in your workplace cannot function without a specific person? Simply recording those tasks on video can serve as future AI training data. Just using a smartphone to film is sufficient. The cost is zero.

2. Keep Track of Information on Robots That Operate ‘Without Teaching’

Watch the trends of robots that operate autonomously with AI control, rather than traditional industrial robots. Specifically, keep an eye on Google DeepMind’s Gemini Robotics, Figure AI (backed by OpenAI), and domestic companies like PFN’s robotics business.

There’s no need to attend trade shows. Watching one demo video on YouTube each month will give you a sense of ‘what can be done now.’

3. Experiment on a Small Scale

There’s no need to wait 3 to 5 years. There are experiments involving ‘AI x the workplace’ that can be done right now.

For example, image recognition for inspection tasks. If you simply have AI evaluate photos taken with a smartphone, you can start with services costing just a few thousand yen per month. It doesn’t have to be perfect. It’s important for the people on the ground to have the experience of ‘having AI perform tasks.’

There’s no need to buy a robot right away. Start by having AI perform one task. That’s where you can begin.

It’s Not That ‘Labor Shortages’ Will Be Solved. The ‘Meaning of Labor’ Will Change

Finally, just one thing.

If you view this technological change solely as a ‘solution to labor shortages,’ you will miss the essence.

What is truly happening is that the boundary between ‘tasks that should be performed by humans’ and ‘tasks that should be performed by machines’ is shifting dramatically.

In the case of chefs, tasks like ‘cutting with a knife’ and ‘stirring a pot’ will shift to robots. So what remains for chefs? Creating menus, selecting ingredients, and adjusting flavors based on customer feedback. In other words, ‘judgment’ and ‘creativity.’

The same applies to manufacturing. Line work will move to robots. What remains for humans is the ability to spot anomalies, think of improvements, and communicate with customers.

The strength of small and medium-sized enterprises lies in their speed, where ‘the president’s judgment is directly linked to the workplace.’ While large corporations are bogged down in bureaucratic processes, small businesses can try, fail, and adjust quickly. In the era of robots + AI, this strength will become even more advantageous.

In three years, a small factory operating with ‘one robot + the president’s judgment’ will achieve the same productivity as a large corporation’s factory. That future is already beginning to come into view.

Will you wait or take action? I believe the answer is clear.