Conclusion

Let’s get straight to the point: “AI is cheap” is a lie.

Data released by Microsoft has completely changed the atmosphere surrounding AI investment.

“In many tasks, AI is more expensive than humans” — this is the very statement from Microsoft, the seller of AI.

Even more shocking is the figure from another survey: 94% of companies whose AI projects are not delivering expected results still continue to spend.

In other words, this means:

• AI is not as cheap as expected
• Results are not being delivered
• But they cannot stop

This is the very essence of the “sunk cost fallacy.” How many department heads in large corporations can stop a project into which they have already invested hundreds of millions of yen by saying, “We are not seeing results”? The answer is clear.

However, it is precisely in this structure that small businesses find their opportunity.

—

Why Do Large Corporations’ AI Investments Become “Expensive”?

First, let’s break down what is meant by “AI is high cost.”

The costs that large corporations invest in AI generally have this structure:

When totaled, it is not uncommon for large corporations’ AI projects to reach 1 billion to 5 billion yen in the first year alone. Moreover, this is the amount before any results are seen.

Why does this happen? Large corporations start with “company-wide implementation,” “infrastructure building,” and “governance.” They spend enormous costs on creating systems before solving on-site issues. As a result, by the time AI starts operating, most of the budget has already been consumed.

And 94% say they “cannot stop.” The larger the amount invested, the heavier the decision to withdraw becomes. This is the reality of AI investment in large corporations.

—

Why Don’t Small Businesses Fall into the Same Trap?

The answer is simple: They simply do not have a budget of several hundred million yen.

This is not a weakness; it is a structural strength.

When small businesses use AI, their options look like this:

While large corporations are spending 100 million yen to create an “AI infrastructure,” small businesses can automate “today’s operations” for just 30,000 yen per month.

Moreover, small businesses do not face the pressure of “cannot stop.” If a 30,000 yen tool does not fit, they can cancel it next month. Even if they spend 3 million yen on a PoC and fail, the company will not be on the brink. “Starting small and withdrawing immediately if it doesn’t work” is a structural advantage of small businesses.

—

What Costs Are Plummeting Behind the Claim That “AI is Expensive”?

This is the main point.

When Microsoft says “AI is more expensive than humans,” it refers to the case of building and operating large-scale AI systems in-house. On the other hand, the cost of “using” AI as a service continues to plummet.

Let’s look at some specifics:

• Translation: Outsourcing specialized translation costs 20 to 30 yen per word. A 10,000-character document would cost 200,000 to 300,000 yen. With ChatGPT, it costs a few hundred yen and takes just a few minutes.
• Meeting Minutes Creation: Outsourcing human transcription for one hour of audio costs 10,000 to 20,000 yen. With AI transcription and summarization, it can be used for a few thousand yen per month.
• Image Production: Requesting a designer costs 50,000 to 100,000 yen per piece. With services like Midjourney, it costs 4,000 yen per month for unlimited use.
• Code Generation: Work that would cost an engineer 5,000 to 10,000 yen per hour can be largely covered by GitHub Copilot for 2,000 yen per month.
• Data Analysis Reports: Requesting from a consulting company costs 1 million to 3 million yen. Feeding company data to Claude can yield initial analyses for just a few thousand yen.

What this means is that “building an AI system is expensive, but using AI-generated outputs is incredibly cheap.”

Large corporations are pouring hundreds of millions of yen into the former. Small businesses should target the latter.

—

The Winning Strategy for Small Businesses Lies in Three “Reversal Structures”

Reversal Structure 1: Speed of Decision-Making Surpasses Technical Capability

The pace of evolution of AI tools is astonishing. Best practices from six months ago are already outdated today. While large corporations are going through the approval process, tools can change two generations.

If the president of a small business says, “Let’s use this,” they can implement it the next day. This speed of decision-making itself is a competitive advantage. A company that can incorporate GPT-4o into its operations the week after its release will clearly have an advantage over a company that sets up a committee to consider its implementation.

Reversal Structure 2: Small Businesses Can Break “Personalization” with AI

The biggest management challenge for small businesses is personalization. There are many tasks that “cannot function without that person.”

However, conversely, the tasks that are personalized are precisely where the impact of systematization through AI is greatest.

Teaching AI the estimation know-how of veteran employees. Feeding ten years of customer interaction history into RAG so that anyone can provide answers of the same quality. This would become a massive project of “cross-departmental data integration” in large corporations, but for small businesses, it can be as simple as “feeding the customer ledger in Excel to AI.”

In fact, in a local manufacturing company, transferring the inspection know-how of veteran craftsmen to AI image recognition was completed in about 800,000 yen and two months. If a major manufacturer were to do the same, just coordinating with the quality control department would take six months.

Reversal Structure 3: The Value of Being Able to Make the Decision to Stop

The fact that 94% cannot stop even after failures — this flipside is a weapon for small businesses.

They can try a 30,000 yen AI tool for three months, and if it does not work, they can stop. They can try another tool. If they cycle through this four times a year, they can conduct four AI experiments for 360,000 yen a year. Meanwhile, large corporations are spending 30 million yen on a single PoC.

“Low failure costs” mean “more opportunities for experimentation.” Organizations that have more opportunities for experimentation ultimately arrive at the correct answer. This is a fundamental principle of innovation.

—

So, What Should Be Done?

Small businesses need to do just three things.

1. Implement one AI tool costing a few thousand yen per month right now.

It can be a meeting minutes AI, ChatGPT, image generation, anything. Start using one today. Thinking can come later.

2. List up the tasks that “only that person can do.”

Tasks that are personalized are where the returns from systematization through AI will be the greatest. Start by making a list.

3. If there are no results in three months, stop without hesitation.

The greatest weapon of small businesses is the ability to “stop.” Do not fall into the sunk cost trap. There is no need to follow the same path as large corporations.

—

The Era of “AI is Expensive” is the Era of Small Businesses

Microsoft has acknowledged that “AI is more expensive than humans.” This is an inconvenient truth for large corporations but good news for small businesses.

Why? While large corporations are bound by “expensive AI” and unable to move, small businesses can feast on “the outputs of cheap AI.”

There is no need to create an AI system. Just use the “cheaper outputs” generated by AI. Translation, images, analysis, code, meeting minutes — things that used to cost hundreds of thousands of yen to outsource are now available for just a few thousand yen per month.

The first to reap this fruit will not be large corporations that take three months for approvals. It will be small businesses that can act the day after the president says, “Let’s do it.”

Now that the illusion that “AI is expensive” has collapsed, it is the perfect time for small businesses to take action.