Teaching AI About All Japanese Companies Correctly.

We endorse April Dream, an initiative to declare dreams on April 1st. This press release expresses the dream of Compalyze Inc.

Compalyze Inc. (Kusatsu City, Shiga Prefecture, Representative Director: Takashi Suzuki), which operates the corporate information database "Compalyze," declares its dream on April 1st, "April Dream."

Try asking AI. It won't be able to answer.

Go ahead and ask ChatGPT, Claude, or Gemini:

"Tell me about manufacturing companies in Shiga Prefecture that have food-related permits and licenses, and whose latest financial results were profitable."

Most likely, you won't get an accurate answer. You might see a list of plausible company names, but the presence of permits and licenses would be speculative, the financial figures would lack sources, and there might even be non-existent companies mixed in.

This isn't because AI is immature. It's because the "correct data" that AI can refer to simply doesn't exist in the first place.

Corporate data "exists." But it's scattered and unreadable.

An enormous amount of information about Japanese corporations actually exists. The problem is that it's scattered in disparate locations, in various formats, and with inconsistent update frequencies.

Registry information is with the Legal Affairs Bureau. Financial statements are in the Official Gazette. Patents and trademarks are with the Patent Office. Permits and licenses are divided among ministries such as the Ministry of Health, Labour and Welfare, the Ministry of Land, Infrastructure, Transport and Tourism, and the Ministry of Agriculture, Forestry and Fisheries. Moreover, registrations at the Legal Affairs Bureau are PDFs, financial statements in the Official Gazette are image data, and permit ledgers from various ministries can be Excel or CSV files; there is no unified format.

In other words, the data isn't "non-existent," but it's "not in a usable state." This is the fundamental reason why AI cannot accurately answer questions about Japanese companies.

Taking on the grimy work that AI cannot do

What Compalyze is diligently working on every day is the task of refining this "unusable data" into "AI-reliable quality."

The content of this work is surprisingly plain and laborious.

It involves using OCR to read financial statements published as PDFs in the Official Gazette and converting the figures into structured data. It distinguishes and categorizes "assignment of claims" and "creditor protection procedures" correctly from registry ledgers. It tracks changes in addresses and trade names to accurately reconcile information for the same legal entity. It identifies dormant companies versus active ones. It links permit and license data scattered across more than five ministries to a single corporate number. And it cross-references financial figures with multiple sources to verify their authenticity.

There are no shortcuts in this work. It's an area where human curation is indispensable and cannot be fully entrusted to AI. Currently, Compalyze's database comprises 90 tables and tens of millions of records. Regardless of whether a company is listed or unlisted, it constitutes a "correct resume" of companies, composed solely of primary information with clear sources: registration, financial results, trends in employee numbers, news, and permits/licenses.

AI is smart. But if it doesn't have data to feed on, it can't answer anything. And if it's fed garbage, it returns garbage. Compalyze is a company that creates the "state where AI can feed on reliable data."

What happens when this data reaches AI?

The data we accumulate daily is not an end in itself. When this data is integrated into the foundation of AI, concrete changes will occur in the Japanese economy and society. That is our true dream.

Local factories will be "discovered" as optimal business partners

Japan has approximately 4 million corporations, most of which are unlisted companies. Many excellent companies with unique technologies and stable management exist in rural areas. However, because information is not organized, they are neither discovered as potential business partners nor noticed by investors.

If corporate information is properly structured and AI can accurately answer when asked, "Tell me about a profitable manufacturing company in Shiga Prefecture that can process precision parts," business opportunities will reach not only large companies in Tokyo but also local factories. Simply by organizing information, the bloodstream of the regional economy will change.

Support will reach companies facing business succession crises

The Small and Medium Enterprise Agency estimated that by 2025, approximately 600,000 companies would face the risk of closure due to a lack of successors. However, there are hardly any external means to identify which companies are in crisis.

If signals such as changes in registered directors, trends in employee numbers, and the status of financial statements are structured, business succession support organizations and M&A brokerage firms can promptly extend a helping hand to companies in need. This might reduce even one of the "silent closures" that have been overlooked due to unorganized data.

Investment decisions will be based on "data," not just "connections"

Information on unlisted companies is always insufficient when investing in startups. Investors are forced to make decisions worth tens to hundreds of millions of yen based on pitch materials and limited meetings. If public information such as chronological changes in registration information, past capital increase histories, and changes in executive composition are centrally organized, the accuracy of due diligence will significantly improve.

If investment decisions based on correct data increase, funds will flow not to "acquaintances' introductions" but to "truly growth-oriented companies." This will lead to the overall healthy development of Japan's startup ecosystem.

Fewer transaction accidents for SMEs

In inter-company transactions, credit assessment of business partners is a matter of life or death. However, major credit investigation services cost hundreds of thousands to millions of yen annually. For small and medium-sized enterprises and freelancers, the reality is that it's "too expensive to use." Cases of starting transactions without sufficient investigation and getting entangled in uncollected accounts receivable or sudden bankruptcies are endless.

If there is a system that allows tracking changes in registration and anomalies in financial information for 980 yen per month, accounting staff at local factories and freelance designers can also detect signals early, such as "that company's executives have changed multiple times recently" or "no financial statements have been published for two years." We want to prevent even one of the accidents that could have been avoided.

The "invisible wall" in employment and job changes will disappear

For listed companies, management status can be understood through IR materials. However, job seekers applying to unlisted companies can hardly know the actual state of the company. "What I heard in the interview was significantly different from the reality after joining"—many people must have had such experiences.

If trends in employee numbers, financial performance, and changes in registration become visible to everyone, job seekers can make judgments based on objective facts such as "this company's employees have tripled in five years" or "it has recently become profitable." Resolving information asymmetry also means protecting workers.

What is needed is not another LLM.

The AI development race is accelerating worldwide. Larger models, faster inference, more parameters. However, we believe that what is truly missing in the AI era is not "another smart model."

What is missing is reliable, accurate corporate data that AI can trust.

No matter how much models evolve, if the data they refer to is inaccurate, the output will remain inaccurate. Conversely, with correct data, even current AI can be sufficiently useful to people. Our job is to painstakingly accumulate that "correct data," one legal entity at a time.

Compalyze is a small startup born in Kusatsu City, Shiga Prefecture. We don't have flashy technical announcements or news of massive funding. However, we are constantly facing corporate information across Japan, structuring primary information one by one, and continuously creating a "correct resume" of companies using only data with clear sources.

Our Dream

When you ask AI, "Tell me about this company," all the information returned has a source.

Whether it's a listed company, an unlisted company, a newly founded startup, or a local factory that has been operating for 50 years.

A "correct resume" of all companies exists, and everyone can access it.

Excellent local companies are discovered, support reaches companies in business succession crises, investments are driven by data, not connections, transaction safety is protected, and workers can choose their future with correct information.

What is needed in the AI era is not another LLM. It's reliable, accurate corporate data that AI can trust.

Compalyze will build that foundation.

This is the dream of Compalyze Inc.

About "April Dream"

April Dream is a project proposed by PR TIMES, where companies, organizations, and local governments declare their "dreams"—not "lies"—on April 1st. We endorse this spirit and are announcing Compalyze's dream.

About "Compalyze"

Compalyze is a corporate database consisting only of primary information with clear sources, such as registration, financial statements, trends in employee numbers, news, and permits/licenses. It visualizes the changes in companies over time, regardless of whether they are listed or unlisted, and all main features are available for 980 yen per month (1,078 yen including tax).

Company Overview

Company Name: Compalyze Inc.

Representative Director: Takashi Suzuki

Location: Kusatsu BASE 9, 2-2 Nishi-Oji-cho, Kusatsu City, Shiga Prefecture 525-0037, Japan

Business Description: Operation of the corporate information database service "Compalyze"

URL: https://www.compalyze.co.jp/