£33 Billion Sitting on the Shelf

The UK’s National Health Service (NHS) has invested approximately £3.3 billion, or about ¥33 billion, in Palantir’s data platform, the Federated Data Platform (FDP).

The NHS announced that 29 hospital trusts are “benefiting from this technology.” But what is the reality? According to an investigation by The Guardian, many trusts are not operating the system properly. Most applications are unused, and one executive even described the NHS’s announcement as “deeply misleading.”

£33 billion spent, and almost none of it utilized.

This is not a story about “AI being unusable.” It’s about “organizations being unable to use AI.”

Why Did the NHS Become a ‘Storing Organization’?

The issue with the NHS is not the quality of the technology; it is a structural problem.

First, there is the distance from the field. The decision to implement the FDP was made centrally. However, it is the local hospital trusts that actually use it, each with different business processes and IT literacy. Just because the central authority declares that it has “implemented” the system does not mean that the field is in a state to “use it.”

Next, there is the ambiguity of purpose. “Integrating data” and “improving efficiency with AI”—these statements are correct. However, if it is not defined at the field level which specific tasks and processes will change, the tools become just an infrastructure cost.

Finally, there is a lack of education and operational design. No matter how excellent the platform is, it is meaningless if users do not know how to use it. In the case of the NHS, there was a critical lack of training for field staff and design for integrating the system into daily operations.

The main reason £33 billion of investment was “shelved” is that there was no design for “who will use it, for what purpose, and how.”

This is not just a story about large organizations. In small and medium-sized enterprises (SMEs), the scenario of “we tried to implement a tool, but no one is using it” happens regularly. The scale may differ, but the structure is exactly the same.

What Did Real Estate Company Heimstaden Do?

There is a contrasting example: the Nordic-based real estate management company Heimstaden. Managing approximately 154,000 residential units, the company has fully integrated OpenAI’s ChatGPT and Codex into its operations, changing the way work is done on the ground.

What specifically happened?

• Automation of Tenant Responses: AI handles inquiries from tenants. A system was established where AI can immediately respond to standard questions—such as contract terms, progress on repair requests, and rules for common areas.
• Democratization of In-House Code Development: By utilizing Codex, even non-engineer employees can customize business tools and create simple applications. The wait time for requests to the IT department has drastically decreased.
• Acceleration of Data Analysis: AI automates revenue analysis for each property and forecasts maintenance costs. Analyses that previously took several days can now be completed in just a few hours.

Notably, Heimstaden is not using a dedicated platform worth hundreds of billions like Palantir. ChatGPT and Codex are tools that can be accessed for a monthly fee of a few thousand to tens of thousands of yen. In other words, they competed not on investment amount but on “designing how to use it.”

Heimstaden’s Chief Digital Officer (CDO) stated that the results of AI implementation are “not just about cost reduction. The biggest change is that employees now feel they can ‘solve problems on their own.'”

This is not just a story of efficiency. It is about changing organizational culture.

The Inverted Structure of £33 Billion vs. Monthly Fees of Tens of Thousands

Let’s take a moment to lay out the numbers.

This comparison shows the fact that the results of AI do not correlate with the amount invested.

In fact, it is the opposite. The more money invested, the slower the decision-making becomes, the greater the distance from the field, and the higher the “risk of not being used.” On the other hand, when the field chooses inexpensive tools, tries them on a small scale, and expands while providing feedback, the adoption rate is overwhelmingly high.

This structure is a tailwind for SMEs.

Why SMEs Can Stand on the ‘Utilization Side’

When large companies implement AI, it often involves lengthy processes such as approvals, PoCs, security reviews, and vendor selection, which can take six months to a year. During that time, tools evolve, and the market changes.

SMEs are different.

• Decision-making is fast. If the CEO says, “Let’s do it,” they can start testing the next day.
• The distance between the field and management is close. It is easier to clarify “who will use it” and “what it will be used for.”
• They can start small. They can begin with ChatGPT Plus at around ¥2,000 per month and expand if it proves effective. There is no need for an initial investment of £33 billion.

In fact, we have seen such examples in the local SMEs we support.

• A manufacturing company with 15 employees: AI was used to semi-automate the creation of estimates. A task that previously took 40 minutes per item was reduced to 10 minutes, saving about 50 hours of labor per month.
• A professional services office with 8 employees: ChatGPT was utilized to draft contracts, reducing outsourcing costs (previously ¥150,000 per month) to almost zero.
• A retail business with 20 employees: AI was tasked with creating product descriptions, tripling the speed of e-commerce listings. Sales increased 1.4 times within three months of implementation.

In all cases, the investment was just a few thousand to tens of thousands of yen per month—less than a fraction of NHS’s £33 billion. Yet, the field is undoubtedly changing.

Three Conditions to Become a ‘Utilizing Organization’

The cases of the NHS, Heimstaden, and SMEs reveal three common conditions for organizations that can fully utilize AI.

1. Start with “What Are We Struggling With?”

Do not start with technology. The starting point should be the field’s pain points, such as “This task is cumbersome” or “This takes too much time.” The NHS began with the technology of a “data integration platform,” while Heimstaden started with the challenge of “spending too much time on tenant responses.” This difference made all the difference.

2. Start Small and Iterate Quickly

Do not implement company-wide from the beginning. Start with one task and one person. If it proves effective, expand to the next department. If it does not work, stop. When you invest £33 billion all at once, you may not even be able to make the decision to stop.

3. Create a ‘Culture of Use’ Rather Than Relying on ‘Capable People’

Instead of depending on a few knowledgeable employees, create an environment where “let’s just try it” becomes the norm. Heimstaden succeeded because they established an environment where non-engineers could use the tools. In SMEs, it is often fastest when the CEO demonstrates how to use it.

So, What Should We Do?

The answer is simple.

Do not buy expensive tools. Fully utilize inexpensive tools.

ChatGPT Plus (approximately ¥3,000 per month), Google Gemini, Claude—practical AI tools are now available for a monthly fee of just a few thousand yen. You do not need £33 billion.

What is needed is:

1. Choose one of the most “cumbersome” tasks in your company today.
2. Apply AI to that task.
3. If you do not see results in 30 minutes, try another task.

That’s all.

The case of the NHS teaches us that “having implemented AI” means nothing. The case of Heimstaden teaches us that only organizations that fully utilize AI can change.

And SMEs already have the conditions to fully utilize it: speed of decision-making, closeness to the field, and the agility to start small.

A CEO who fully utilizes a tool costing ¥3,000 a month is far stronger than a £33 billion giant system.

The issue is not the technology. It is whether to use it or let it sit. That is all.