How to connect two systems that have no interface, and why the third one costs less than the second
Connecting two systems without an interface is faster and cheaper than a classic integration project. What matters is how the effort grows once a third and fourth system join.
This article was generated by AI. Labelled in accordance with Article 50 of the EU AI Act. Responsible for publication: Sophera Consulting.
The most expensive answer to two systems that do not talk to each other is replacing one of them. The second most expensive is a person who retypes every transaction from one into the other. Both answers come from a time when an integration was only possible through a purpose-built interface.
Why systems without an interface can still be connected
An agent does not need an API to connect two systems. It needs a way in and a way out. A report as a file, a mailbox, an export function, a screen a person can fill in: whatever a clerk reads and types, an agent can read and type.
That removes the condition which used to block these projects. The vendor of the legacy system had to cooperate. Now it is enough that the system releases anything at all.
The limit is still hard. If a system gives up neither a file nor access, an agent will not help either. And a file export is not a substitute for an interface, it is a cost of its own, because formats change and nobody announces it.
The arithmetic of the third system
The interesting part starts with the third system, and the reason is arithmetic.
Connect systems directly to each other and n systems need n times (n minus 1) divided by 2 connections. Two systems need one. Three need three. Four need six, five need ten, ten need 45. Every new system costs as many new connections as there are systems already.
Route the handover through a middle instead and every system needs exactly one link to that middle. Ten systems cost ten links. The tenth system costs one link rather than nine.
That is why this route gets cheaper with every further system instead of more expensive. The effort grows linearly, where direct connections grow quadratically. It is the whole reason such a chain can be continued until every system involved is attached.
What the agent in the middle does
It takes a transaction out of the source system, puts it into the shape the target system expects, and creates it there. The work sits in the translation: the same field is named differently on each side, a value exists only on one side, an entry cannot be matched with certainty.
That last case is where an agent differs from a rigid interface. An interface either stops or writes something wrong. An agent can flag the case and put it up for review instead of guessing.
Two things have to be settled. First a direction: which system owns which field. Synchronising both directions at once is almost always a mistake. Second, the point where a person approves, for as long as the transaction moves money or dates.
What makes this fast and affordable
No new product appears that has to be licensed and maintained. What appears is a routine that operates the systems you already run.
How fast depends on the process. For order intake from customer emails, meaning the route from the mailbox into the target system, the build takes a few hours, and with one person from the department the process is live within a day. Sophera Consulting builds this for a fixed price and without a subscription, which also means the third and fourth link is commissioned on its own rather than billed as a standing fee per connection.
Where to start
Find the one handover somebody performs by hand today, and write down where the data comes from and where it goes. That single route is the middle everything else attaches to.
In the free Automation Check we look at which of your systems release anything, which route pays off first, and what attaching further systems costs after that.
This article was created with the help of AI.