Don’t Discard That SOAP API. AI Agents Transform ’20-Year-Old Business Systems’ into the Ultimate Weapon

“Our system is too old to be of any use” — I want to ask business owners who think this.

Have you forgotten that there are 20 years’ worth of data lying dormant in that old system?

When they got a quote for replacement costs, it was 30 million yen. When considering a cloud migration, the annual running costs alone were 5 million yen. Many companies end up postponing the decision, thinking, “Maybe it’s fine as it is.” However, the situation is beginning to change. AI agents are now able to connect directly to those troublesome legacy SOAP APIs.

There is no need to discard the system. There is no need to replace it. By simply placing an AI agent ‘on top’ of the old system, the dormant data comes to life.

This is not just a story for large corporations. In fact, it is particularly relevant for small and medium-sized enterprises (SMEs) that lack replacement budgets.

—

SOAP API is Not a ‘Dead Technology.’ It’s Just That No One Knew How to Talk to It

When hearing about SOAP APIs, many engineers frown. They are seen as verbose, XML-based, with thick specifications, and are often regarded as ‘fossils’ compared to modern REST APIs or GraphQL.

But consider this: Many core systems operating in Japanese SMEs were built in the early 2000s. Order management, inventory management, customer management — many of the interfaces produced by these systems are SOAP-based. And they are packed with 10 to 20 years’ worth of business data.

The problem was not that “SOAP APIs are unusable.” The problem was that there were no longer people who could ‘talk’ to the SOAP APIs.

The development vendors from that time have already withdrawn. There are specifications, but no one can read them. WSDL files remain, but there are no personnel in-house who can interpret them and write the code to call them.

This is where AI agents come in. In recent proof-of-concept projects, AI agents have successfully auto-analyzed WSDL files (the specification definitions for SOAP APIs) and autonomously assembled the necessary requests to extract data from the old systems. There is no need for humans to decipher the specifications. The agents become the ‘interpreters’ for the old systems.

In one case, an AI agent connected to the SOAP API of an order management system that had been operational for 20 years, enabling real-time analysis of past transaction data. The additional system development costs were almost zero. They simply used the existing API as is.

A 30 million yen replacement or an almost zero yen AI agent connection. It is clear which option is more realistic for SMEs.

—

A Time When You Can ‘Ask the Database’ in Natural Language

Now that we can extract data from legacy systems, the next barrier is “how to use that data.”

Traditionally, analyzing business data required personnel who could write SQL. How many companies in the SME sector have employees who can write SQL? In reality, most can only manage to “manually drop it into Excel and aggregate it with pivot tables.”

Now, NL-to-SQL (automatic conversion from natural language to SQL) technology is rapidly approaching practical levels.

Specifically, AI can convert Japanese inquiries such as “Tell me the top 10 customers from last month’s sales” or “Which products are below the reorder point?” directly into accurate SQL queries and return answers from the database.

The NL2SQL benchmark released by DevRev has shown that the latest architectures can handle complex schemas (structures with over 100 tables) with high accuracy. While this is a research-level discussion, practical applications are undoubtedly progressing.

The implication for SMEs is simple. The cost of ‘hiring engineers to view data’ will disappear.

Instead of a data analyst costing 500,000 yen a month, AI tools costing a few thousand to tens of thousands of yen a month can perform the same job. Moreover, they do so 24/7 with zero wait time.

The excuse of “We don’t have data personnel” will no longer hold water.

—

AI Solves the Problem of ‘Not Knowing What Systems Exist’

Another issue that SMEs often overlook is the problem of not knowing what systems or tools they have, with no one grasping the overall picture.

Core systems, accounting software, attendance management, customer management, custom Excel spreadsheets… Each department uses them in isolation, with no comprehensive list even existing. When trying to introduce new tools, they cannot determine overlaps or integration possibilities with existing systems.

Here, the automatic construction of a software catalog using coding agents is noteworthy.

In one proof-of-concept experiment, a prototype was completed in a three-day hackathon. The coding agent scanned the internal system environment and automatically listed the software, APIs, and databases in operation, visualizing their connection and dependency relationships.

What previously cost 2 to 5 million yen for an IT asset inventory through external consultants could potentially be done by the agent in just a few days and at almost no cost.

If you understand your company’s ‘possessions,’ you can make informed decisions about what to connect and what to discard. If you can make those decisions, you can avoid unnecessary investments. For SMEs, this is a subtle yet extremely significant advantage.

—

The Structural Reason Why ‘Old’ Becomes ‘Strong’

Let’s summarize the discussion so far.

1. Legacy SOAP API × AI Agent → No need for replacement, 20 years of data can be utilized immediately.
2. NL-to-SQL → Even without knowing SQL, you can query the database in natural language.
3. Coding Agent × Software Catalog → Automatically visualize your IT assets.

What these three have in common is the idea of ‘using existing assets as they are.’ Instead of buying new systems, it’s about changing the ‘usage’ of what is currently available.

And here lies a reversal structure unique to SMEs.

Large corporations have too many legacy systems, making integration take years. Governance barriers, inter-departmental adjustments, security reviews — in a world where it takes three months just to get approval for connecting one AI agent.

On the other hand, SMEs have fewer systems and faster decision-making. “Let’s try it next week” is a feasible approach. Speed is the greatest weapon for SMEs, and AI agents further accelerate that speed.

—

So, What Should We Do?

Putting aside the complex discussions, here are three things you can start doing today.

1. Search for Your Company’s Legacy System API Specifications
WSDL files, API specifications, documents left by development vendors. Check if they are hidden away in the back of your drawers. Just having them can open the door to connecting AI agents.

2. List Up ‘What You Want to Ask’
Even when using NL-to-SQL, it’s meaningless if you’re not clear about ‘what you want to ask the data.’ Have employees in the field list five things they manually check each month. This will directly serve as candidates for automation.

3. Start Small
There’s no need to think about full company-wide implementation. Connect just one API, one database, one inquiry. Start with that alone using the AI agent. If it works well, you can expand it. If it fails, the loss is almost zero.

—

Think ‘Connect’ Before You ‘Discard’

Japanese SMEs have business systems that have been operational for 20 or 30 years. They are not ‘negative legacies.’ They are assets of business data accumulated over 20 years that cannot be bought with money.

Thanks to the evolution of AI agents, the cost of utilizing those assets has dramatically decreased. A 30 million yen replacement could potentially be resolved with almost zero yen in connections. Even without knowing SQL, you can ask data in Japanese. The overall picture of your company’s systems can also be organized by the agent automatically.

Before making the judgment to ‘discard’ old systems, I hope you will first try to ‘connect.’

Your company’s strongest weapon is already inside the server room.