Conclusion

Let’s get straight to the point: It’s still too early to cut staff with AI.

According to a recent report from a U.S. HR research firm, 55% of companies that reduced their workforce by implementing AI express regret. Meanwhile, global investments in AI are projected to exceed $200 billion (approximately 30 trillion yen) in 2024 alone. Despite their regrets, investments continue unabated. What accounts for this contradiction?

The answer is simple: What AI can do and what it is acceptable to reduce human resources for are entirely different matters. Companies that have confused these two concepts are now facing painful consequences.

For small and medium-sized enterprises (SMEs) in rural areas, this is not just someone else’s problem. Rather, it presents a rare opportunity to witness the failures of large corporations firsthand. Let’s break down what is happening across three layers: cost, human resources, and infrastructure.

—

Layer 1: Cost—Where Has It Become Cheaper and Where Has It Become More Expensive?

“A task that used to cost 3 million yen has been reduced to 50,000 yen thanks to AI.” Such stories do exist. In areas like routine data entry, image classification, and initial customer inquiries, the cost-cutting power of AI is real.

However, the problem lies beyond that.

The “invisible costs” that arise after implementing AI are causing companies distress. Specifically:

• Maintenance and operational costs of AI systems: Even a SaaS costing several tens of thousands of yen per month can incur additional expenses due to customization and integration with internal data. It’s not uncommon for annual costs to balloon to two or three times the initial estimate.
• Costs of maintaining accuracy: AI deteriorates if left unattended. If data trends change, retraining is necessary, incurring costs each time.
• “Filling the gaps” costs after cutting staff: Tasks that AI could not replace remain, placing a heavier burden on the remaining employees. As a result, turnover increases, and recruitment costs soar.

In the U.S., Ars Technica reports that about 40% of AI data centers may not be completed on schedule. This situation affects even infrastructure investments worth billions of dollars.

However, SMEs must not misunderstand: there is no need to build their own data centers. Cloud-based AI services can be accessed for a few thousand yen per month. The issue lies in the design of “what to use AI for,” not the scale of infrastructure.

While large corporations spend billions to create their own AI frameworks and struggle with their operation, SMEs can utilize AI with the same performance for under 10,000 yen per month through a single API. This asymmetry is the weapon of SMEs.

—

Layer 2: Human Resources—Is It True That “There Are No AI Personnel”?

“We can’t find AI talent.” This concern is shared by both large and small companies. But what exactly do we mean by “AI personnel”?

Many companies are looking for engineers who can design machine learning models from scratch, individuals with salaries ranging from 10 to 20 million yen. There are only a few thousand such individuals across Japan. It’s unlikely that rural SMEs can win in this recruitment competition.

However, the situation fundamentally changed after 2024.

With APIs for large language models (LLMs) like ChatGPT and Claude, no-code/low-code AI tools, and AI features even integrated into Google Sheets, having “people who can create AI” is not necessary; having “people who can use AI” is sufficient for SMEs to effectively utilize AI.

In fact, in a local manufacturing company we support (with 30 employees), part-time administrative staff have automated the generation of estimates using ChatGPT. The time required for each estimate was reduced from 40 minutes to just 3 minutes, saving approximately 500 hours of labor annually. There are no specialized AI engineers involved.

So what is “AI fatigue”?

It’s not a problem of talent shortage, but rather a misalignment of expectations. Companies think, “If we implement AI, everything will go smoothly,” and then become exhausted by the gap between expectations and reality. AI startups emerge one after another, claiming, “Our AI can do this and that.” They try it out, but it doesn’t meet their expectations. They move on to the next tool, and this cycle is the essence of “AI fatigue.”

What’s needed is not “new AI tools” but a blueprint for “where to integrate AI into our business.” The tools can be chosen later.

—

Layer 3: Infrastructure—Is the Energy Issue Relevant to SMEs?

The power consumption of AI data centers is staggering. A single large data center can consume as much electricity as a mid-sized city. In the U.S., opposition from local residents against data center construction is growing, and the environmental impact is becoming a serious issue.

However, let’s be honest: This is not something that rural SMEs need to worry about directly.

Why? Because when SMEs use AI, it’s almost 100% through the cloud. The electricity issues of data centers are problems that companies like Microsoft, Google, and Amazon, which operate them, should solve. The costs borne by SMEs as users do not directly correlate (aside from indirect impacts like rising electricity prices).

Instead, the infrastructure that SMEs should focus on is their internal data foundation.

• Customer data is scattered across Excel files.
• Past estimates and order histories exist only in the minds of the responsible staff.
• Daily reports and documentation are managed in various formats.

AI “only works if there is data.” Conversely, if the data is organized, the utilization of AI can advance rapidly. Delays in data center construction are less critical than organizing your own Google Sheets, which is a far more significant bottleneck for AI utilization.

—

So, What Should We Do?

Large corporations regret their AI layoffs. Fatigue towards AI startups is spreading. Data center construction is delayed.

When faced with these news stories, do you think, “AI is still too early,” or do you see it as a chance, thinking, “Now that large corporations are stumbling, it’s our opportunity”? This is the dividing line.

What SMEs should do can be narrowed down to three points.

1. Don’t implement AI to cut staff. Implement it to free up human time.

The essence of regret over AI layoffs is that “too many people were cut.” AI is not omnipotent. Judgment, negotiation, and relationship-building with customers—these are tasks that only humans can perform. Automating routine tasks with AI and redirecting the freed-up time to “work that only humans can do” is the correct approach.

2. Don’t hire expensive “AI personnel.” Teach your current employees how to use AI.

Teaching five existing employees how to use AI tools is far more cost-effective than hiring one AI engineer with a salary of 15 million yen. If the cost of AI tools is 2,000 yen per month per person for five people, that’s 10,000 yen per month. If it doubles efficiency for an annual cost of 120,000 yen, there’s no reason not to do it.

3. Organize your company’s data before searching for tools.

Eighty percent of the success or failure of AI utilization depends on “data organization.” No matter how excellent the AI, garbage in means garbage out. Start with an inventory of customer lists, sales data, and business flows. It may seem mundane, but this is the most reliable first step.

—

The Nature of the Contradiction

“55% of AI layoffs regret it” and “AI investment is at an all-time high.” This contradiction is not a contradiction at all.

The value of AI itself is genuine. It’s just that companies that miscalculated the design of “what to change with AI” are the ones regretting it.

Large corporations implemented AI to “reduce staff.” That’s why they regret it. SMEs should implement AI to “empower people.” Even with the same technology, the results can change 180 degrees based on how it’s used.

SMEs can learn for free from the failed experiments of large corporations that spent billions. There has never been a more fortunate time than this.

First, let’s give it a try. You can start for just 10,000 yen a month. Even if you fail, it won’t be a fatal blow. But if you do nothing for three years, that will indeed become a fatal blow.