Title

33 Billion Yen AI Left Unused by 29 Organizations: Don’t Overlook This Structure Where a Small Factory Recoups Its Investment First with 50,000 Yen a Month

Main Text

The UK’s National Health Service (NHS) has implemented an AI data platform through a contract worth £330 million (approximately 33 billion yen) with Palantir. What has been the result? Twenty-nine hospital trusts have reported that they are “barely using it.”

33 billion yen spent, yet it remains unused.

This is not just someone else’s problem. The structure in which large organizations fail to implement AI and small to medium-sized enterprises (SMEs) can achieve results with an investment of 50,000 yen a month are two sides of the same coin.

—

Reasons Why 33 Billion Yen is “Sitting on the Shelf”

The NHS officially claims that these hospital trusts are “benefiting from” Palantir’s platform. However, the voices from the field tell a different story. Multiple trusts have stated that they are “not operating the system” and “not using the application.” There is a clear gap between the NHS’s claims and the realities on the ground.

Why does this happen? There are three reasons.

1. Too Many Layers of Decision-Making
In a massive organization like the NHS, introducing a single tool to the field requires passing through multiple layers of decision-making, including approvals from department heads, validation by the IT department, security reviews, budget approvals, and training plan development. The speed of AI evolution outpaces the speed of organizational decision-making.

2. Unclear Who’s Job Will Change
Even if AI tools are introduced, if it is not defined where in the workflow they will be applied, the field will continue with traditional methods. In large organizations, business processes are intricately intertwined, making it extremely difficult to change just one aspect.

3. Ambiguous Responsibility
Who is responsible for the ROI on an investment of 33 billion yen? Is it the procurement department, the IT department, or the field managers? When responsibility is dispersed, the worst-case scenario arises: “it’s not being used, but no one is stopping it.”

This is not just an issue for the NHS. In Japan’s large corporations, there are numerous cases where AI projects worth tens of millions of yen end up as mere “Proof of Concept (PoC).” Structurally, the larger the organization, the higher the hurdle to “make use of” AI.

—

Reasons Why SMEs Can Recoup Their Investment First with 50,000 Yen a Month

So, what about SMEs? Here, we see a reversal of structure.

Let’s do some calculations.

Consider an SME with 10 employees, where 2 administrative staff spend a total of 40 hours a month on invoice processing, data entry, and responding to inquiries. If we calculate at 2,000 yen per hour per person, that amounts to 160,000 yen a month.

By combining ChatGPT Team (about 5,000 yen per person per month) with automation tools like Zapier (about 5,000 yen per month), we can implement a solution for about 20,000 yen a month. This would automate initial responses to inquiries, delegate data extraction from invoices to AI, and semi-automate input tasks.

Assuming we can reduce 50% of those 40 hours, that would mean a reduction of 20 hours. This translates to a labor cost reduction of 80,000 yen per month. The investment is 20,000 yen a month, resulting in a profit of 60,000 yen from the first month. This leads to an annual savings effect of 720,000 yen.

While large corporations hold onto “unused AI” worth 33 billion yen, SMEs can turn a profit from the first month with an investment of just 20,000 yen. This asymmetry represents the greatest opportunity of our time.

To provide a more realistic example, consider a local manufacturing company that takes 30 minutes to create a quote. If they handle 60 quotes a month, that totals 30 hours. By implementing an AI system that learns from past quote data to automatically generate quotes, they could potentially reduce the time to 5 minutes per quote. This would save 25 hours a month, resulting in over 50,000 yen in labor savings.

The key point is that “being small allows for speed.” Once the CEO decides to “do it,” they can test it the next day. The approval process has only one layer. Even if it fails, the loss is only a few tens of thousands of yen. This agility is something that large organizations, which must go through a 33 billion yen procurement process, cannot replicate.

—

The Evolution of AI Increases the Value of “Agility”

There is another trend to keep in mind here: the pace of AI evolution itself.

Dario Amodei, CEO of Anthropic, recently warned that AI is approaching the stage of “Recursive Self-Improvement.” This is the phase where AI improves its own code and accelerates its performance.

What this means is that the performance of AI tools could be completely different in six months.

For large organizations, this is a nightmare. A system that took two years to implement could become obsolete the moment it goes live. On the other hand, SMEs can simply switch to a monthly subscription SaaS. This month it’s ChatGPT, next month it’s Claude, and in six months it might be a tool whose name we don’t even know yet. SMEs, which can use AI as a variable cost rather than a fixed asset, are better positioned to adapt to this pace of evolution.

Moreover, the wave of “Agent AI” is also coming. This is a system where AI autonomously performs tasks without human instructions. Tasks like replying to emails, scheduling adjustments, and making inventory ordering decisions could become “automatically completed.”

For large corporations to implement Agent AI, they would need to revise security policies, negotiate with labor unions, and conduct integration tests with existing systems. SMEs can simply start with, “Let’s try it out in the accounting department for invoice processing.”

—

So, What Should We Do?

The lessons from the NHS’s failure are clear.

“Going big to change everything” will fail. “Starting small to change one task” is the correct approach.

Here are three things SMEs should do starting today.

1. Identify the “most time-consuming task”
There’s no need for company-wide DX. Whether it’s invoice processing, quote creation, responding to inquiries, or taking meeting minutes—just pick one.

2. Experiment within a budget of 10,000 to 30,000 yen per month
ChatGPT (from 3,000 yen), Claude (from 3,000 yen), Zapier (from 3,000 yen). This combination is sufficient to get started. No initial investment is required. If it doesn’t work, just cancel the subscription the following month.

3. Measure only “how many hours were reduced”
Don’t complicate ROI calculations. Just ask, “How many hours did this task take last month? How many hours did it take this month?” That’s all you need. The reduction in hours multiplied by the hourly wage will directly translate to investment returns.

—

33 Billion Yen vs 20,000 Yen a Month: Which Will Prevail?

The NHS case demonstrates that the value of AI is determined not by “how much was invested” but by “how quickly it was integrated into the field operations.”

Even with an investment of 33 billion yen, if 29 organizations do not use it, its value is zero. Conversely, even with just 20,000 yen a month, if it eliminates 20 hours of labor from the first month, that represents a guaranteed return.

There’s no need to mimic large corporations. SMEs have their own way of fighting. Being small is not a weakness; it is, at this moment, the greatest weapon.

Tomorrow morning, think of the most tedious task in your company. That will be the first job for AI.