##靴を売るのをやめたら株価が6倍になった。これ、おかしくないか？

Do you remember the shoe brand Allbirds? It took the world by storm with its sustainable sneakers and went public in 2021 with a market capitalization of over $4 billion. Since then, it has plummeted. Sales have nearly halved from 2022 to 2025, and ultimately, the brand and its shoe assets were sold for just $39 million, about 1% of its IPO value.

And here comes the bizarre part.

CEO Joe Vernachio renamed the remaining publicly traded shell to “NewBird AI” and announced a pivot to an AI company offering GPU-as-a-Service (GPUaaS). There are no shoe inventories or stores left, only the shell of a public company and a $50 million fundraising plan.

The stock price surged by 600%.

Let’s take a step back. This is not a story of “success in business through AI.” This is a story of the stock market reacting to the banner of AI. The stock price is soaring at a stage where there are no sales, customers, or products. If small business owners see this and think, “We should pivot to AI too,” that would be a completely misguided lesson.

SDLは逆に「AIコード全面禁止」を宣言した

At the same time, an entirely opposite movement is occurring. The open-source game library SDL (Simple DirectMedia Layer) has declared a complete ban on commits (code submissions) generated by AI.

The reasons are simple.

• The licensing of AI-generated code is unclear, posing legal risks.
• The volume of code generated cannot be adequately reviewed by humans.
• Trust within the developer community is at risk.

SDL is a project with over 25 years of history, forming the backbone of thousands of games and applications. For them, “code reliability” is a lifeline. What if AI-written code gets mixed in and licensing issues arise later? The entire project would face legal risks. This judgment is understandable.

However, this should not be taken at face value.

“The complete ban on AI code” does not mean that avoiding AI is the correct answer. It signifies a judgment that, given the current quality of AI code generation and legal frameworks, the risks are too high for their use cases. In other words, this is a discussion about risk management, not a denial of technology.

「AI丸乗り」も「AI全拒否」も、中小企業の正解ではない

When we juxtapose these two cases, a structure becomes apparent.

Allbirds (NewBird AI): A strategy to attract stock market money under the banner of AI. The reality is yet to come. The movement is driven more by expectations than substance.

SDL: A strategy to accurately assess the risks of AI and eliminate it for now. Cautious, but also missing out on opportunities.

Both may be rational in their respective contexts. However, it is dangerous for local small businesses to mimic either of them.

The trick of raising funds using the guise of a public company is not something small businesses can pull off. On the other hand, continuing to fight with only traditional methods by completely banning AI is a path that will gradually lead to losing in cost competition.

So where is the correct answer for small businesses?

中小企業のAI活用は「コスト構造を変える」一点に絞れ

To put it simply, what small businesses should do with AI is not “new ventures” or “complete bans,” but to change the cost structure of existing operations.

Let’s talk numbers specifically.

• Creating estimates: What took a person 30 minutes can be done by AI in 5 minutes. If there are 100 estimates a month, that’s a reduction of 41 hours. At an hourly wage of 2000 yen, that’s 82,000 yen a month. About 1 million yen a year.
• Responding to inquiries: By assigning FAQ responses to an AI chatbot, 60-70% of inquiries can be automated. This saves the cost of one part-time employee, around 1.5 to 2 million yen a year.
• Creating minutes and reports: AI can transcribe and summarize recordings of meetings. A 30-minute task can be reduced to 3 minutes. If there are 5 meetings a week, that’s a reduction of 9 hours a month.
• Screening recruitment documents: It takes a full day for a person to review 100 applications. With AI screening, it takes just 1 hour.

These are not flashy stories. They are not about stock prices soaring by 600%. But achieving annual cost reductions of 3 to 5 million yen can be done with AI tools costing just a few thousand to tens of thousands of yen a month. This is the essence of AI for small businesses.

「AIを使う」と「AIに賭ける」はまったく違う

What Allbirds did was “bet on AI.” They fully committed their business to AI. This is akin to gambling.

What small businesses should do is “use AI.” Keep the current business as it is, but lower the costs of running the business with AI.

This distinction is critically important.

For example, let’s consider a local manufacturing business.

• Core business: Metal processing. This cannot be replaced by AI. The skills of artisans and equipment are competitive advantages.
• Areas to change with AI: Automatically classify and draft replies for order emails. Image inspection for quality control. Automatically generate monthly reports.

The strengths of the core business remain intact. However, the costs of supporting “indirect operations” can be halved. If a company with 5 employees reduces indirect operations by 80 hours a month, that time can be redirected to the core business. Alternatively, reduced overtime could lead to lower employee turnover. Recruitment costs decrease.

This is the form of AI utilization that small businesses can achieve. Large corporations are too big to make such nimble improvements. They have to go through approvals, select vendors, and conduct PoCs six months later. Small businesses can start next week.

SDLの判断から学ぶべきは「何にAIを使わないか」を決めること

There is another lesson to be learned from SDL’s complete ban on AI.

It is to clarify “areas where AI should not be used.”

For small businesses, this would look like the following delineation.

• Areas where AI can be used: Internal document creation, data aggregation, schedule management, FAQ responses.
• Areas where AI can be used but must be confirmed by humans: Customer-facing emails, estimates, proposals.
• Areas where AI should not be used: Final management decisions, direct face-to-face communication that affects trust with customers, final verification of legal documents.

Without these delineations, implementing AI on a whim can lead to accidents. There are real cases where an AI-generated estimate contained incorrect numbers and was sent to a customer as is.

Conversely, if the delineation is clear, AI can be used confidently. If internal meeting minutes are somewhat inaccurate, it’s not a major issue. But if an invoice to a customer is incorrect, that could be catastrophic. Using risk levels to differentiate is the essence of what small businesses should learn from SDL’s judgment.

で、結局どうすればいいのか

You only need to do three things.

1. This week, choose one indirect operation that takes the most time in your company.
Minutes, email replies, data entry, anything is fine. Just one.

2. Apply an AI tool to that operation.
ChatGPT, Claude, or Gemini will do. It costs about 2000 to 3000 yen a month. It’s not a 3 million yen system implementation. Even if it fails, the loss is just 3000 yen.

3. Determine “areas where AI should not be used.”
Things that directly involve customers, those with legal risks, and those that affect brand trust. These should be handled by humans.

There’s no need to bet the entire company on AI like Allbirds. Nor is it necessary to completely reject AI like SDL. Start with 3000 yen a month and reclaim 10 hours a month. That is the first step for small businesses in utilizing AI.

Don’t be swayed by the flashy story of a 600% stock surge. Grounded cost reductions are the true essence of AI utilization for small businesses.