Conclusion

Let’s get straight to the point. AI does not generate profits; it reduces costs.

Many business leaders believe that investing in AI will increase profits. However, the reality is different.

According to economist Daron Acemoglu’s analysis, the boost to the U.S. GDP from AI adoption will be a mere 0.93% to 1.16% over the next decade. In the face of trillions of yen in investment, the macroeconomic returns are surprisingly small. Even if large corporations pour billions into AI, they are not seeing commensurate top-line growth.

So, does this mean AI is meaningless?

No. The intrinsic value of AI lies not in ‘generating profits’ but in ‘eliminating costs.’

And the small and medium enterprises (SMEs) that benefit most directly from cost reductions are not the large corporations. Why? Large companies face enormous ‘implementation costs’ such as organizational changes, security assessments, vendor selection, and internal approvals when adopting AI. SMEs can start moving as soon as the owner says, “Let’s do it.” This speed difference translates directly into profit differences.

“A 3 Million Yen Job Becomes 50,000 Yen” Is Happening in Reality

Let’s discuss some concrete examples.

For instance, consider the tasks that a local manufacturing company has been outsourcing:

• Manual Creation and Translation: Outsourcing costs 800,000 to 1,500,000 yen annually → If internalized using generative AI, the effective cost is almost zero. API costs are just a few thousand yen per month.
• Simple Development of Internal Systems: Outsourcing costs per project range from 1,000,000 to 3,000,000 yen → By using open-source coding models, a prototype can be created with just prompts and reviews. Outsourcing costs drop to less than one-tenth.
• Inquiry Responses: Part-time labor costs 2,000,000 to 3,000,000 yen annually → Automating first responses with AI chatbots can reduce labor costs by 40% to 60%. This saves 800,000 to 1,800,000 yen annually.
• Minutes and Reports Creation: Employee work hours range from 20 to 30 hours per month → With speech recognition and summarization AI, this can be reduced to 2 to 3 hours per month. Over 200 hours of labor can be eliminated annually, equating to 400,000 yen at an hourly wage of 2,000 yen.

When these are totaled, even a small to medium enterprise with 20 employees can save 3,000,000 to 5,000,000 yen annually. Business owners understand how difficult it is to increase sales by 3,000,000 yen. Cost reduction translates directly into profit.

Why Do Large Corporations End Up with “Zero Profit”?

Let’s clarify the structure here.

The reasons large corporations cannot generate profits from AI are clear:

1. Implementation Costs Are Too High
To adopt AI, large corporations need to establish security requirements, integrate with existing systems, conduct internal training, and build governance structures. Relying on system integrators can cost tens of millions to hundreds of millions of yen. Even if the cost of AI itself decreases, the ‘cost of implementing AI’ does not.

2. Competitors Are Also Implementing Simultaneously
When large corporations adopt AI, the benefits of efficiency are offset by competition. If everyone has the same weapon, no one can win. This is known in economics as the “Red Queen Effect”—running continuously but staying in the same place.

3. Organizational Rigidity
Even if operations are streamlined with AI, surplus personnel cannot be immediately redeployed. The reduced labor hours do not translate into profit but are absorbed within the organization.

On the other hand, what about SMEs?

• Implementation costs are nearly zero. ChatGPT costs 3,000 yen per month, and Claude costs 3,000 yen per month. Open-source models are free.
• Competing SMEs are not yet using AI. Only the companies that act first can capture the difference.
• The owner is the decision-maker. “We’ll start using this tomorrow” is all it takes. No internal approvals or committees are needed.

This structure reveals the paradox of “AI’s economic benefits are zero, but only SMEs can win.”

Open Source Models Are Becoming a “Free Arsenal”

From 2024 to 2025, the performance of open-source AI models has dramatically improved.

For example, Meta’s Llama 3.1, Mistral’s Mixtral, and various models from China. These can be used for free and operate on local PCs. Since data is not sent to the cloud, there is no risk of data leakage.

Particularly noteworthy is the evolution of coding-specific models and speech models.

Using coding models, internal tools like “Excel macros” can be created without hiring engineers. Automated notifications for inventory management, automatic aggregation of order data, and automatic summarization of daily reports—tools that are “convenient but not worth outsourcing” can be created with just a few lines of prompts.

Outsourcing would cost 500,000 to 1,000,000 yen per item. Creating it yourself costs virtually zero. This price difference becomes the very competitiveness of SMEs.

The same goes for speech models. With the release of small speech models that operate locally by NVIDIA, it has become technically possible to delegate first responses in phone support to AI. 24-hour availability, zero waiting time. Even local construction firms and retail stores can provide customer support comparable to large corporations for just a few thousand yen per month.

“So, What Should We Do Ultimately?”

That’s enough of the abstract discussion. Let’s narrow down what small and medium business owners should do starting tomorrow into three actions.

1. First, Create an “Outsourcing Cost List”

Write down how much your company pays for outsourcing annually. Translation, design, system development, manual creation, data entry, inquiry responses. Choose one item from this list that can be replaced by AI and try it out next week. It doesn’t have to be perfect. A score of 70 is sufficient.

2. Provide a 3,000 Yen AI Tool to All Employees

Distribute ChatGPT Plus or Claude Pro to all employees. For 10 employees, that’s 30,000 yen per month. That’s 360,000 yen annually. If this results in 10 hours of work being eliminated per person each month, that’s a reduction of 1,200 hours annually. At an hourly wage of 2,000 yen, that’s a return of 2,400,000 yen. The ROI is 6.6 times. There is no more cost-effective investment than this.

3. Don’t Appoint an “AI Person”; Create a State Where “Everyone Uses AI”

The worst-case scenario is to appoint one person as the AI representative and call it a day. This leads to dependency; if that person leaves, it’s over. Instead, create an environment where everyone uses AI in their daily work. Simply sharing “What did you try with AI this week?” in morning meetings is enough. Make it a system. Make it a culture.

What Happens After Costs Decrease

Finally, let’s discuss what happens further down the line.

As cost reductions from AI progress, the “break-even point” for SMEs will decrease. When the break-even point lowers, previously unprofitable businesses can become viable.

To be specific—

• An e-commerce site that couldn’t operate with monthly sales of 500,000 yen can become profitable due to reduced operating costs.
• A sole practitioner in a professional services firm can handle the workload of three people with the help of an AI assistant.
• A local inn can directly attract inbound customers with a multilingual AI chatbot.

What was once only possible for large corporations becomes achievable for SMEs. This is the true value of AI. It doesn’t matter if profits are zero. The mere fact that costs decrease becomes the greatest weapon for SMEs.

The question is whether businesses will recognize this structure and take action.

Technology won’t wait. If competitors lower their costs first, they will win in price competition. While you leave AI as something “not well understood,” the company next door has increased its gross profit margin by 10%. That era has already arrived.

**Let’s stop thinking about “What can AI do?” and start calculating “How much can we save with AI?” Only those companies that can capture that difference will survive the next decade.