Training Costs: How Much Are You Spending Annually?

The training costs for small and medium-sized enterprises (SMEs) are often invisible yet steadily pressure management.

According to a survey by the Small and Medium Enterprise Agency, companies with around 50 employees spend an average of 800,000 to 1.5 million yen annually on training-related expenses. This includes costs for external instructors (100,000 to 200,000 yen per session, 4 to 6 times a year), materials creation (50,000 to 100,000 yen several times a year), venue costs, and, most importantly, the opportunity cost of employees’ time spent on training.

This structure is about to fundamentally change with AI.

A platform called InfinityMock has established a system that automatically generates mock tests for 110 different exams and qualifications using AI. Furthermore, research is emerging that shows the learning effects can be enhanced when AI agents “teach each other,” as well as studies indicating that custom-designed AI chatbots demonstrate higher educational effectiveness than generic chatbots.

Will the cost of employee training truly become zero? To conclude, it will not become zero. However, reducing “1.2 million yen to 200,000 yen” is quite realistic.

The Shock of Automatically Generating Exam Questions for 110 Subjects

What InfinityMock has achieved is not merely “random generation of questions.”

By analyzing the question trends, difficulty distribution, and frequently asked themes of each exam, it generates a large number of mock questions of quality close to actual exams. The 110 exams covered include IT-related qualifications (such as Fundamental Information Technology Engineer and AWS Certification), accounting qualifications (such as bookkeeping and tax accountant subjects), and industry-specific qualifications.

Let’s reflect on what companies have traditionally done to support employees in obtaining qualifications:

• Purchasing commercially available question books: 2,000 to 4,000 yen each multiplied by the number of employees
• Attending external preparatory courses: 30,000 to 100,000 yen per person
• Creating mock exams internally: several days to weeks of labor for the responsible person

If you want to have 10 employees obtain IT-related qualifications, it would cost 40,000 yen for question books and 300,000 to 1 million yen for preparatory courses. If creating mock exams internally, assuming a labor cost of 3,000 yen per hour and 20 hours to create questions for one subject, that would be 60,000 yen. For 10 subjects, that totals 600,000 yen.

With AI-generated content, this “materials creation cost” can be nearly zero. The platform usage fee is expected to range from several thousand to tens of thousands of yen per month, totaling several tens of thousands to over a hundred thousand yen annually. This is less than one-tenth of traditional costs.

The Inverted Idea of “Learning by Teaching AI”

There is a noteworthy study in the field of education: “Learn by Teaching Your AI Agent Teammate” — a method where humans learn by teaching an AI agent.

This implements the long-known educational principle that “teaching is the best way to learn” using AI. Employees teach the AI agent their business knowledge, and the AI acts as a “student” by asking questions. Employees must organize and deepen their knowledge to answer these questions.

Traditional training has been a one-way street from “instructor to participant.” Participants passively receive information, and understanding varies greatly among individuals. However, in the process of teaching AI, employees must actively structure their knowledge. This proactivity significantly enhances learning effectiveness.

Practically, consider this usage: a new employee “teaches” the company’s operational manual to the AI. The AI asks questions about parts it does not understand. The new employee then seeks answers from seniors or revisits the manual. As a result, they gain a much deeper understanding than if they had simply “read” the manual.

Custom AI Chatbots vs. Generic Chatbots — The Difference in Effectiveness is Clear

When using AI for training, the most crucial decision is whether to use a generic ChatGPT as is or a customized chatbot tailored to the company’s training content.

Research results are clear. Custom AI chatbots that incorporate the Socratic method (a technique that guides thinking through questions rather than directly providing answers) significantly improved participants’ problem-solving abilities compared to generic chatbots.

On the other hand, when using generic chatbots, participants are more likely to engage in the behavior of “asking the AI for answers.” This is known as “cognitive offloading.” They outsource thinking to AI, which hampers the development of their own critical thinking skills.

This difference is obvious when considering the purpose of training. The goal of training is not just to “know the correct answer” but to “develop the ability to arrive at the correct answer.” Simply asking a generic AI for answers does not achieve this goal.

The cost of building a custom chatbot has significantly decreased. By using OpenAI’s GPTs or the custom instructions feature of ChatGPT, basic custom bots can be created in a few hours without programming. Even if outsourced, it would cost around 100,000 to 300,000 yen. Compared to annual external instructor fees, this can be recouped in the first year.

How Cost Structures Will Change

Let’s compare the traditional training costs and the costs after utilizing AI, assuming a manufacturing company with 50 employees.

Traditional Annual Training Costs:

• External instructor fees (6 times a year): 600,000 to 1.2 million yen
• Materials and question book purchases: 100,000 to 200,000 yen
• Venue and equipment costs: 50,000 to 100,000 yen
• Internal staff preparation labor (100 hours annually): equivalent to 300,000 yen
• Total: 1.05 million to 1.8 million yen

Annual Training Costs After AI Utilization:

• AI question generation platform usage fee: 60,000 to 120,000 yen
• Custom AI chatbot development and operation: 100,000 to 200,000 yen (first year; 50,000 to 100,000 yen from the second year)
• External instructor fees (2 times a year, for advanced content only): 200,000 to 400,000 yen
• Internal staff operational labor (30 hours annually): equivalent to 90,000 yen
• Total: 450,000 to 810,000 yen (first year), 400,000 to 720,000 yen (from the second year)

The reduction amounts to 600,000 to 1 million yen annually, with a reduction rate of about 50 to 60%. While it will not become “zero,” it can be reduced to less than half.

Reasons It Won’t Become Zero — There Are “Trainings Only Humans Can Do”

To be honest, training costs will never reach zero. This is because there are definitely trainings that cannot be replaced by AI.

On-the-job training (OJT), team building, leadership training, and mental health training are all fundamentally about human interaction and cannot be substituted by AI. Additionally, training on the latest industry trends or legal changes requires the insights of experts.

AI excels at “inputting knowledge and confirming retention.” Generating exam questions, supporting repetitive learning, and acquiring foundational knowledge are areas where AI can take over, allowing human instructors to focus on “trainings that only humans can do.”

This is not a story about “AI taking away human jobs.” It is about “AI taking over mundane tasks so that humans can concentrate on more valuable work.”

Things You Can Start Today

There is no need for a large-scale system implementation. Here are some things you can start today.

1. Try creating mock questions about your business using ChatGPT. Simply input, “Please create 10 comprehension check questions for new employees regarding our XX business.” Check the quality of the output to see the AI’s capabilities for yourself.

2. Visualize your current training costs. Calculate the total annual cost, including not only outsourcing fees but also the preparation time of internal staff and the time commitment of participants. Understand where there is room for reduction in numerical terms.

3. Experimentally introduce AI as a “supplementary material” in your next training session. Instead of a complete switch, start by adding AI-generated practice questions to traditional training. Gradually expand based on participant feedback.

Reducing training costs directly correlates with the competitiveness of SMEs. Saving 600,000 to 1 million yen annually is equivalent to one employee’s bonus. This capital can be redirected towards improving employee salaries or investing in equipment.

AI will not bring training costs to zero. However, it will clearly distinguish between “necessary costs” and “costs that can be reduced.” Delegate the reducible parts to AI and concentrate investment on the non-reducible parts. This is the smart training strategy for SMEs.