Conclusion First

Saying “We’re done with AI” is not a shame.

Xbox’s new CEO, Asha Sharma, announced the halt of the AI project “Copilot.” At the same time, the EU is facing criticism for its €2 billion (approximately ¥320 billion) investment in AI computing, deemed a “waste.”

Both large corporations and governments have failed to recoup their investments in AI.

Interpreting this as “AI was a failure” is a mistake. The accurate interpretation should be, “AI investments without a clear purpose fail, regardless of scale.” What truly matters for SMEs is whether they have the criteria to decide not just “when to start” but “when to stop.”

—

What Did Xbox Get Wrong?

First, let’s clarify the facts.

Xbox’s Copilot was a project aimed at integrating AI assistants into the gaming experience, helping players strategize in games.

Sharma explained the reason for the halt: “Xbox needs to deepen its relationship with the community and alleviate friction between players and developers.”

In short, what users wanted was not an AI assistant. The dissatisfaction that players felt was not due to the absence of AI, but rather a lack of connection with the community and friction with developers. The problem was misidentified.

This is not an issue unique to Xbox. The moment “implementing AI” becomes the goal, the same issue can arise in any company.

—

Where Did the EU’s €2 Billion Go?

The EU invested €2 billion (approximately ¥320 billion) in AI computing infrastructure starting in 2024, aiming to enhance Europe’s competitiveness in AI. However, there has been a backlash from the field, with criticisms of “unclear usage” and “lack of visible results.”

That’s ¥320 billion. The median annual revenue of a small Japanese company is about ¥150 million. This amount corresponds to the annual revenue of over 2,000 companies, yet it remains in a state of “unclear usage.”

The lesson to be drawn from this is simple: the size of the investment does not correlate with results. In fact, the larger the amount, the more pressure there is to “spend it all,” leading to vague objectives.

I have seen many cases where a small monthly AI investment of ¥50,000 in SMEs directly correlates to tangible results.

—

Criteria for SMEs to Decide to “Stop AI”

So how should SMEs determine whether to “continue or stop”?

You can temporarily set aside textbook discussions about “ROI,” “market trends,” and “scalability.” There are only three criteria that are practically useful.

Criterion 1: Has “someone’s job become easier” within three months?

After introducing AI tools for three months, is there someone in the company saying, “I can’t do without this”? If not, it’s okay to stop.

In numerical terms, how many hours of work have been reduced per week? How many mistakes have been decreased per month? If you cannot answer these questions with specific numbers, that AI investment is not effective.

For example, in a manufacturing company, AI was introduced for creating estimates. A task that previously took 40 minutes per estimate was reduced to 8 minutes. With 50 estimates a month, that results in a reduction of 26 hours monthly. At an hourly wage of ¥2,000, that translates to a saving of ¥52,000 per month. If the monthly cost of the AI tool is ¥30,000, the net gain is ¥22,000. It’s small, but it is definitely effective.

You should focus on whether such “small certainties” accumulate.

Criterion 2: Will anyone be inconvenienced if we revert to “without AI”?

This is the most brutally honest yet accurate test. Try stopping the AI tool for a while.

If no one is inconvenienced, then that AI was unnecessary. If voices arise saying, “We need it back,” then that AI has become established.

Large corporations do not conduct this test. Once implemented, they must report that “results are being achieved.” SMEs do not have that political pressure. This is why they can make honest judgments. This is the strength of SMEs.

Criterion 3: Is the monthly cost exceeding that of “one part-time worker”?

Many failures in AI investments for SMEs stem from cases where “initial costs are too high.”

For example, an initial development cost of ¥3 million and a monthly operating cost of ¥200,000 is more expensive than hiring a part-time worker. Moreover, a part-time worker can often be more flexible in various situations.

The current market for AI tools looks like this: ChatGPT Plus costs around ¥3,000 per month, Claude Pro is also around ¥3,000, and specialized SaaS ranges from ¥10,000 to ¥50,000 per month. No-code automation tools can cost anywhere from a few thousand to tens of thousands of yen monthly.

In other words, there are countless AI initiatives that can be tested for under ¥50,000 per month. Before listening to vendors proposing a ¥3 million development project, first try something for ¥50,000 for three months. If it proves effective, then expand it. If not, stop. This order is correct.

—

Companies That Can Make the Decision to Stop Will Ultimately Win in AI

It is easy to laugh at Xbox’s decision as a failure in AI. However, what we should really focus on is the fact that they were able to make the decision to stop.

Many large corporations cannot stop AI projects once they have started. “It was approved in a management meeting,” “the budget was secured,” and “a press release was issued.” Their pride and sunk costs prevent them from withdrawing. As a result, they continue to pour tens of millions of yen annually into ineffective AI projects.

The same structure applies to the EU’s €2 billion. Politically, they cannot say “we’re stopping.”

SMEs do not have this constraint. If the CEO thinks, “This is pointless,” they can stop the next day. The speed of decision-making is the greatest weapon of SMEs.

And this weapon can be used not only for “the speed of starting” but also for “the speed of stopping.” In fact, the speed of stopping is more important.

—

So, What Should We Do?

Here are three key points for SMEs to keep in mind when utilizing AI:

1. Test for three months for under ¥50,000 per month. Instead of development projects, apply existing AI tools to your operations.
2. After three months, confirm with numbers “whose work, what tasks, and how many hours were reduced.” If you cannot answer, stop.
3. Do not consider stopping as shameful. Both Xbox and the EU have stopped. If your company stops, no one will be inconvenienced.

While large corporations spend hundreds of millions of yen saying, “We’re done with AI,” SMEs are using tools costing ¥30,000 a month to reduce estimate creation time by one-fifth. This reversal structure is the reality of the current AI era.

What matters is not whether you are using AI, but whether the challenges in front of you are being resolved.

Companies that can honestly confront this question will ultimately win, regardless of their size.