“Isn’t This Irrelevant to Us?”

Until just two years ago, it was “normal” for a business system to cost 3 million yen. You would ask a vendor to define the requirements, go through multiple customizations, and repeat testing. It took six months to deliver. Additional modifications could cost another 1 million yen. For small and medium enterprises in regional areas, IT systems were synonymous with being “expensive, slow, and ultimately unusable.”

That common understanding is now fundamentally breaking down.

With open-source AI agents, no-code sandbox environments, and automatic knowledge graph generation tools, we have entered an era where you can build a “reasonably functioning business system” for under 50,000 yen a month. 3 million yen has become 50,000 yen. The cost is now one-sixtieth. We should seriously consider the risks of continuing to say, “It’s still too early for us.”

Why Has It Become One-Sixtieth? — Three Technical Backgrounds

There are three changes behind this drastic cost reduction.

1. Explosive Increase in Open-Source AI Agents

Starting in late 2024, there has been a rapid increase in business-specific open-source AI agents. For example, using a framework like Feynman, you can automate customer inquiries, internal FAQs, and simple workflow management by feeding in your company’s manuals and business data. Previously, outsourcing such systems to a system integrator would cost at least 1.5 to 2 million yen. With open-source, the tools themselves are free. All you need is the cost of a cloud server (a few thousand to tens of thousands of yen per month) and the time for initial setup.

2. Tools That Organize Information by Just Dropping It In

Tools like CacheZero allow you to simply drop documents into a folder, transforming their contents into an interactive wiki. This enables search and query capabilities. The time previously lost to questions like “Where is that document?” or “Only Mr. XX knows this” can be reduced to almost zero.

In a company with 10 employees, if each person spends 30 minutes a day searching for documents, at an hourly wage of 1,500 yen, that amounts to about 1.12 million yen lost annually to searching. The implementation cost of this tool is nearly zero. There’s no need to calculate ROI.

3. Visualization of Business Processes Through Automatic Knowledge Graph Generation

Using tools like Graphify, you can automatically visualize the relationships in business processes. You can see who is responsible for what and where information flows. The “business flow diagrams” that consultants used to create for 1 million yen can now be auto-generated by simply inputting data.

The combination of these three factors has created a situation where “we can build business systems ourselves without relying on vendors.”

The Real Problem Is Not the “Tools” but the “Personalization”

However, there is one important caveat here. Just implementing tools will not change anything.

The biggest reason for inefficiency in small and medium enterprises is not the lack of systems. It is that the business exists only in the minds of specific individuals. This is what is known as personalization. If a veteran employee leaves, the business may come to a halt. Manuals exist, but they are three years old and no one has updated them. In such a state, even if you introduce AI tools, there is no data for the AI to consume.

In other words, the first thing to do is not to “introduce AI” but to “prepare the business in a form that AI can consume.”

Five Steps to Implement “Systematization”

Here are the specific steps to follow.

Step 1: Inventory of Business Operations (1-2 Weeks)

First, have all employees write down “the tasks they perform every day.” Excel is fine. Task name, time required, frequency, necessary information, and criteria for judgment. This alone will clearly show which operations are personalized. The cost is zero. All that is needed is the management’s directive.

Step 2: Digitalization of Data (2-4 Weeks)

Paper manuals, know-how in the heads of veteran employees, and implicit rules. Convert these into text data. It doesn’t have to be perfect. Just recording with voice input and transcribing it is sufficient. Using free transcription tools like Whisper, one hour of recording can be converted into text in just ten minutes.

Step 3: Selection and Implementation of Tools (1-2 Weeks)

Choose appropriate tools based on the nature of the business. For customer inquiries, use an AI chatbot. For document management, use a wiki tool like CacheZero. For visualizing business flows, use knowledge graph tools. The important thing here is to not aim for perfection from the start. First, try it with one operation, and if it works well, expand it.

Step 4: Implementation of Automation (2-4 Weeks)

Feed the data organized in Step 2 into the selected tools. Start automating “tasks that don’t need to be done by humans,” such as automated responses to standard inquiries, automatic aggregation of daily reports, and automatic updates of inventory data.

Step 5: Operation and Improvement (Ongoing)

It doesn’t end with implementation. Set up a weekly 15-minute session to reflect on whether “this tool is being used properly.” If it’s not being used, investigate the reasons. If the data is outdated, update it. This steady cycle will make the system a “living entity.”

Cost Comparison: Traditional vs. AI-Utilizing

Let’s compare with specific numbers, assuming a manufacturing company with 15 employees.

Looking at annual costs, the traditional model costs 3.6-6.8 million yen in the first year. The AI-utilizing model costs 120,000-700,000 yen. The difference can be over 6 million yen, equivalent to the annual salary of one employee.

The Question of “Is This Enough?”

Many business owners think, “We are managing fine with our current methods.” But is that really true?

What is the average age of veteran employees? When that person leaves, will the business continue to run? How many months does it take for a new employee to become productive? What is the opportunity loss during that time?

In a world where a 3 million yen system has become 50,000 yen, will you continue to run your business on “paper and the memories of veterans”? When competitors adopt this system, can you win on price? Can you win on speed?

The answer is clear.

Future Points of Interest

By late 2025, the accuracy of open-source AI agents will further improve. Support for the Japanese language is also rapidly advancing. While we are still at a stage where “only a limited number of people can utilize it,” in six months, the UI will improve, and even employees who are not IT-savvy will be able to operate it.

The longer you wait, the easier it will become to implement. However, the costs and opportunities lost during that waiting period will not come back.

The first step is to conduct an inventory of business operations. It costs zero, and there is no risk. It can be done starting today. There is no reason not to do it.