The Era of Paying 3 Million Yen for System Integration is Coming to an End

When small and medium-sized enterprises (SMEs) try to connect two or three cloud-based business systems, estimates come in quickly. “300,000 yen for API integration” or “500,000 yen including customization.” The tasks they want to accomplish are simple, yet due to the systems not being able to “speak the same language,” a human intermediary is required, design documents are created, and tests are run.

This structure may fundamentally change.

Currently, technology is rapidly being developed that allows AI agents to converse directly with each other using a “common protocol (communication standard).” Without human intervention, agents can convey intentions, exchange data, and complete tasks. If this becomes a reality, the majority of the system integration costs that SMEs have been paying could disappear.

We will focus on three movements:

• MPAC (Multi-Principal Agent Coordination Protocol)—A mechanism for agents from different organizations to cooperate without conflict.
• Multi-Agent Interoperability Protocol—A system that allows agents created by different vendors to communicate in a common language.
• XBPP (eXtensible Business Payment Protocol)—A mechanism for agents to autonomously complete payment transactions.

Let’s examine what each of these changes means from the perspective of SMEs.

—

“Our Inventory Management” and “Our Order Management” Start Talking on Their Own

First, let’s discuss MPAC (Multi-Principal Agent Coordination Protocol).

Until now, business system integration has fundamentally been something that “humans design.” Data from System A is passed to this field in System B. Conditional branching is like this. In case of an error, it’s like this. SIers provide estimates in the hundreds of thousands of yen for this design and implementation.

What MPAC aims for is a world where this “design” itself is entrusted to agents. It consists of five layers—session management, intention clarification, operation, conflict management, and governance—where agents from different organizations declare “what they want to do” and automatically adjust in case of conflicts.

Imagine this concretely.

For example, a small manufacturing company. Inventory management is on Company A’s cloud, order management is on Company B’s SaaS, and accounting is done via freee. Currently, to connect these three, they either manually output and import CSV files or spend 1 to 3 million yen annually on API integration. If data inconsistencies arise, staff members manually reconcile them. It’s common for companies to spend 20 hours a month on this task.

In the MPAC world, AI agents embedded in each system exchange events like “an order has been received,” “inventory has been allocated,” and “sales have been recorded” using a common protocol. There is no need for humans to write design documents. The agents understand each other’s intentions, avoid conflicts, and complete the processing.

The possibility exists that the 3 million yen in SI costs could drop to zero. The 20 hours spent on reconciliation could also disappear. This is not just “efficiency”; it’s the “disappearance of structure.”

—

“We Can’t Connect Because the Vendors Are Different” Will Become Obsolete

Next, let’s talk about the Multi-Agent Interoperability Protocol.

Here, a more deep-rooted problem is addressed: vendor lock-in.

When SMEs choose systems, they always ask, “Can tools from Company A and Company B integrate?” If the answer is “no,” they have no choice but to lean towards one. As a result, they end up using tools that do not actually fit their business needs. Alternatively, they may introduce intermediary systems just for integration, further inflating costs.

The Multi-Agent Interoperability Protocol solves this problem at the “agent layer.” If each vendor’s system supports this protocol, agents can communicate in a common language. Just as HTTP became a common foundation in the web world allowing any browser to communicate with any server, a common foundation will emerge for agents as well.

When this becomes widespread, what will change for SMEs?

They will be able to freely choose the optimal tools.

This vendor has the cheapest inventory management. This one is easier for customer management. Freee is best for accounting. Even if they choose different tools, the agents will connect them automatically. They will be liberated from vendor lock-in.

This has a greater impact on SMEs than on large corporations. Large companies can pay SIers to connect anything. SMEs cannot do that, so they have had to compromise. The standardization of protocols will break this asymmetry.

—

“Send the Invoice, Wait for Payment, and Reconcile” Will Be Done Automatically

The third is XBPP (eXtensible Business Payment Protocol). This may have the most significant impact on the ground.

Ask an accounting staff member at an SME, “How many hours does it take for month-end invoicing?” You might hear answers like 10 hours, 20 hours, or even two full days. Creating invoices, sending them out, confirming payments, reconciling, and following up—this entire process repeats every month.

XBPP is a protocol that allows AI agents to autonomously handle everything from the occurrence of a transaction to the completion of payment.

For example, the following flow could be realized: The order agent detects “delivery completed.” A notification is sent to the purchasing agent of the other company. Inspection is conducted automatically, and the payment agent executes the remittance process immediately based on XBPP. On the company’s side, payment confirmation and reconciliation are completed automatically.

The only task left for humans is to approve exception handling.

The 20 hours of accounting work could be reduced to 2 hours. The payment cycle could even change from “end-of-month closing with payment at the end of the following month” to “same-day payment after delivery.” For SMEs, improving cash flow is a matter of life and death. Companies struggling with cash flow and facing bankruptcy while still in the black may be saved by this system.

—

So, What Should We Do Now?

To be honest, these protocols are still in the process of standardization. MPAC is in the research stage, the interoperability protocol is in the specification drafting phase, and XBPP is also in the proof-of-concept phase. This is not something that can be used starting next month.

However, the direction is clear. The assumption that “system integration is something designed by humans” is collapsing.

So, what should SMEs do now? There are three things:

1. Get into the habit of choosing “API-compatible” tools.
When protocols become widespread, legacy systems that do not even have APIs will be left behind. When introducing new tools, always check whether they have an API and if they are open to external integration. Just doing this will drastically change future migration costs.

2. “Verbalize” business flows.
To delegate tasks to agents, it is necessary to clarify “what is done in what order.” Black-box operations that are too personalized cannot be handed over to AI agents. Now is the time to take stock of business flows and verbalize them. This is a fundamental aspect of management, not just in the age of AI.

3. Experiment with small AI agents.
With ChatGPT’s GPTs, Dify’s workflows, and Make (formerly Integromat) automation, “mini-agents” can still be created. For a few thousand yen a month, automate inquiry responses, summarize daily reports, or automatically transcribe order data. Companies that accumulate such small experiments will be able to scale rapidly when protocols are established.

—

The Real Competition Will Come After “Protocols Become the Norm”

I also want to mention a structural aspect.

When protocols are standardized and system integration costs approach zero, what will happen? “Being connected” will no longer be a differentiator. Since everyone can connect.

At that time, the differentiation will come from “what to do once connected.” In an environment where data flows automatically, what decisions will be made? What customer experiences will be created? At what speed will actions be taken?

This is an opportunity for SMEs. Large corporations make slow decisions. They have five levels of approval. SMEs can change things starting tomorrow if the president says, “Let’s do it.” They can look at the data collected automatically by agents, make decisions on the spot, and act immediately. This speed is something large corporations cannot replicate.

The field of “system integration costs”—a battle of endurance—will disappear, transforming into a field where “speed of judgment” and “resolution on the ground” are the competitive factors. This is the arena where SMEs excel the most.

The standardization of protocols may seem like a technical issue, but it is actually about competitive structure. The assumption that “large corporations win because they are large” is about to crumble once again.

The gap between companies that start preparing now and those that say, “It’s still too early” will likely become decisive in about 2 to 3 years.