Will Old Code Hold Back Japan’s AI Era?

Japan’s modernization still runs on old code

A salary appears in a Japanese bank account. An insurance claim is paid. A factory orders parts. Railways, logistics companies, utilities, local governments, hospitals, trading houses, and manufacturers update their records. On the surface, smartphone apps glow, AI chats, and cloud platforms seem to run the world.

Behind that surface, however, business systems written decades ago continue to work quietly. COBOL, mainframes, UNIX servers, proprietary specifications, complex batch processes, and programs that have branched through years of modification still carry the core of Japanese enterprise operations.

Fujitsu and IBM Japan’s June 17, 2026 modernization collaboration is aimed at this hidden core. The companies will combine Fujitsu’s source-code conversion solution, Fujitsu PROGRESSION, with IBM’s AI agent-driven development platform, IBM Bob, to rewrite COBOL programs running on Fujitsu mainframes and UNIX servers into Java, then use AI-assisted correction and refactoring after conversion.

This is not a flashy AI story. It is not a chatbot, a robot, or a spacecraft. But for Japan’s digital transformation, it may be one of the most practical and important stories of all. Japan’s future depends not only on new AI, but on how old business systems are understood, preserved, and transformed.

Why legacy systems cannot simply be thrown away

The phrase “legacy system” sounds cold. Old. Slow. Complex. Difficult to maintain. All of that can be true. But from the perspective of an enterprise, a legacy system is not merely an old machine.

It contains the business itself: billing rules, credit decisions, inventory flows, tax treatment, exception handling, customer promises, and decades of small operational adjustments. Some of this knowledge is no longer in manuals. It lives in code. Employees retire, but business logic remains inside programs.

That is why modernization is not a simple relocation. It is not like moving furniture from an old house into a new one. It is more like replacing the pipes and electrical lines under an old city without turning off the city.

You cannot stop bank settlements. You cannot stop insurance payments. You cannot stop factory production planning. You cannot stop local government services. Japanese companies fear legacy systems not because they are old, but because they are too important to break.

The 2025 Digital Cliff is not past — it is underfoot

In its 2018 DX Report, Japan’s Ministry of Economy, Trade and Industry warned of the “2025 Digital Cliff,” the risk that Japanese companies would lose digital competitiveness and suffer serious economic losses if they could not solve problems in existing systems.

2025 has passed. The problem has not. The cliff was not a calendar date; it remains under companies’ feet. In May 2025, METI published a comprehensive report from its Legacy Systems Modernization Committee, again warning that legacy systems hinder the adoption of advanced digital technology and may lead to a decline in Japan’s industrial competitiveness.

The report focuses not only on technical upgrades, but also management awareness, IT asset visibility, autonomy of information-systems departments, collaboration with business units, vendor relationships, human-resources development, and industry-level ecosystems. The problem is not only programming language. It is also decision-making and corporate culture.

To fix old systems, companies may have to change old organizational habits too.

The historical weight of Fujitsu and IBM working together

The combination of Fujitsu and IBM Japan carries historical weight. Fujitsu completed the FACOM 100 in 1954, regarded as Japan’s first practical relay-type automatic computer. It was a major milestone in Japanese computing history, emerging from relay technology used in telephone switching systems.

IBM, meanwhile, has long been deeply involved in Japan’s enterprise computing market, from punched-card machines and accounting systems to mainframes and corporate information systems. IBM’s technology and business model shaped the way many Japanese organizations processed information.

After the war, Fujitsu, Hitachi, NEC, IBM Japan, and others competed in large-scale computer markets. Compatible machines, proprietary systems, operating environments, peripherals, software, and customer bases formed a huge industry supporting Japanese enterprise computing.

Seen against that history, the new collaboration is not just a vendor partnership. Two lineages that once helped build Japan’s enterprise computing landscape are now working together to hand old systems to the next generation.

Why COBOL is still alive

COBOL was born in 1959. As its name — Common Business-Oriented Language — suggests, it was designed for commercial and administrative processing. It aimed for readability and was well-suited to accounting, transactions, reports, and batch processing.

To modern eyes, COBOL looks old. But old does not mean irrelevant. In banking, insurance, government, manufacturing, and distribution, long-running code can become a kind of accumulated trust. It has processed business correctly for decades.

The problem with COBOL is not only the language. The problem is that the people are disappearing. Engineers who know old specifications retire. Fewer people understand entire codebases. Design documents are outdated. Modification histories are complex. External system connections are tangled. A system may still run, but no one can explain all of it.

That is black-boxing. A system is not safe merely because it runs. It can be dangerous because it runs and cannot easily be changed.

The practical path: Fujitsu PROGRESSION and IBM Bob

The center of the collaboration is rewriting COBOL into Java and refactoring the converted code. Fujitsu PROGRESSION converts COBOL programs running on Fujitsu mainframes and UNIX servers into languages suitable for open environments such as Java. Fujitsu’s experience is intended to preserve consistency in business logic and compatibility with existing specifications while reducing migration risk.

IBM Bob then enters the process. As an AI agent-driven enterprise development platform, IBM Bob will support code correction and refactoring after COBOL-to-Java conversion. The goal is to make labor-intensive processes such as business-logic validation, testing, and restructuring more efficient and higher quality.

The important point is that this is not a claim that AI will solve everything automatically. It is a path that uses AI while preserving human judgment and business understanding. In legacy modernization, code conversion is not enough. The converted structure must be maintainable, extensible, testable, and understandable to the business.

Rewriting is the doorway. Refactoring is the redesign.

AI is not magic; it is a translator

AI-driven legacy modernization is often overhyped. Feed old COBOL into a tool, get clean Java, move to cloud, cut costs, and complete digital transformation. If it were that simple, the 2025 Digital Cliff would never have existed.

In reality, AI is not a magician. It is more like a skilled translator. It reads an old language, produces a new language, organizes structure, creates design documents, suggests tests, and visualizes relationships humans may miss. But judging whether the translation is correct still requires people who understand the business.

Japanese enterprise systems often contain many non-standard business rules: customer-specific exceptions, regulatory workarounds, old commercial customs, department-level operations, and manual-process interfaces. AI can read code, but it does not automatically understand why every exception exists.

That is why AI-era modernization should not be about making engineers unnecessary. It should be about increasing the quality of conversation among engineers, business units, and management.

The mainframe is not the villain

In modernization debates, the mainframe is often treated like the villain. Old, closed, expensive, difficult to connect to cloud. Those criticisms can be fair. But mainframes have supported core business for decades. In areas that require stable high-volume transaction processing, they earned trust by not stopping.

The real problem is not the mere existence of mainframes. The problem is the inability to change systems with business needs, the inability to use data effectively, the shortage of skilled people, unclear cost structures, and management’s limited visibility into IT assets.

Modernization therefore cannot be reduced to a slogan of getting off the mainframe. Companies must decide what to keep, what to migrate, what to rewrite, what to separate, what to expose through APIs, and what to move to cloud. The answer differs by company, function, and risk.

Visibility is the real challenge for Japanese companies

As METI’s report emphasizes, the first step in legacy modernization is visibility. What is running where? Who uses it? Which business processes does it support? What data flows through it? Which parts are truly obsolete, and which still work well?

In many companies, this map has been lost. Individual employees understand fragments. Vendors understand pieces. Old design documents exist but no longer match reality. Executives may know total IT spending, but not which systems create competitiveness and which systems hold the company back.

AI-based code analysis and design-document generation may be powerful here. They can help visualize dependencies, process flows, data fields, exception conditions, and duplicated logic. If a system cannot be seen, it cannot be changed.

The human cliff is arriving too

The legacy-system problem is also a talent problem. People who can write COBOL, operate mainframes, understand old reporting workflows, know batch sequences, and remember abnormal-case procedures are aging.

Young engineers often move toward cloud, AI, Python, JavaScript, containers, and mobile applications. Few choose to specialize in COBOL and mainframes. But society’s most important systems still depend on them. That creates a mismatch.

The significance of Fujitsu and IBM using AI for modernization is not only efficiency. It is also about filling the skills gap, converting tacit knowledge into explicit knowledge, and building a bridge that allows younger engineers to understand older systems.

The goal is not to erase legacy, but to inherit it

The word legacy has two meanings. It means something old, and it means an inheritance. Japan’s core enterprise systems are both. They are old, complex, and difficult to maintain. They also contain decades of business knowledge and customer handling.

The purpose of modernization is not to throw away that inheritance. It is to preserve it in a form that can be used in the future. Use AI. Read the code. Convert to Java. Restructure. Connect to cloud and data platforms. This is not destruction. It is translation, restoration, and redesign.

Japan’s DX will not move forward only through glamorous new ventures. It will require the messier work of changing old systems while they continue to run. The Fujitsu-IBM collaboration faces that messy reality directly.

AI-era competitiveness depends on whether old systems can change

In the generative AI era, companies will compete on whether they can use data. Customer data, inventory data, payment data, manufacturing data, contract data, medical data, logistics data. But if that data is trapped inside old core systems, AI cannot fully help.

To use AI, companies need data foundations. To build data foundations, they need to understand business-system structures. To understand those structures, they must read old code and old operations. Japan’s AI competitiveness therefore depends not only on the newest models, but also on how it handles old COBOL.

Fujitsu and IBM Japan’s collaboration strikes a quiet but important target. Before companies can talk about using AI, they must become companies that AI can actually help.

Japan’s future is not only inside futuristic screens. It is also inside black cabinets, old code, and batch processes nobody wanted to touch. Reading that world, preserving it, and handing it to the next generation may be where Japan’s digital transformation truly begins.