An AI application that works within one team reaches the boundary of that team at some point. A summary made by marketing must be readable by sales. A signal that arises in customer service must reach product. Once that happens, it is no longer a question about the model or the prompt, but about how departments deal with each other when something moves between them that did not exist before.
The chat window experience gives the impression that AI is something one person uses to do one task. Beneath that window lies a different reality, certainly once the work extends across more than one department. Anyone setting up a chain in which steps follow each other discovers that every handover between departments is a point where something can get stuck: a format that doesn't match, a responsibility that no one claims, a system that one department has and the other doesn't.
Coordination between departments requires three things that are independent of which model is used.
Departments often use the same words for different things. A lead means something different at marketing than at sales. A complaint is classified differently by customer service than by product. As long as AI applications remain within one department, that difference doesn't stand out. Once output from one department becomes input for another, the difference becomes a malfunction. This applies especially to work that leans on text and conversation: anyone who reads and summarizes documents for another department must know which terms mean something different there, and anyone who logs and follows up on conversations must know who takes over the follow-up once the conversation touches another department.
An application that stops at the boundary of a team remains within the mandate of that team leader. Once the application crosses multiple departments, that mandate is no longer sufficient. There must be someone who can decide across departments: who takes which step, who is responsible if something goes wrong, who determines whether the application gets adjusted. Without that person, every department keeps pointing to the other. This is one of the reasons why a demo without an owner delivers nothing: a demo shows that something can work, but says nothing about who safeguards the collaboration between departments once the real work begins.
Departments typically build their own systems, with their own assumptions about where data resides and who may access it. Coordination between departments requires that these systems can talk to each other, or that there is a layer in between that can do so. This is not a matter of willingness; it is a matter of what is already in place. An organization in which every department has built its own separate system faces a different task than an organization with one shared foundation. That difference becomes visible once work engages agents that actually carry out steps: what you can do with agents that carry out work depends on whether that agent has the same access and the same rules at department A as at department B.
A pilot that runs in the corner of one team never shows where coordination between departments gets stuck, because that coordination is never tested. The pilot appears to succeed, and the question of what happens once a second department is connected remains unanswered until the moment it goes wrong. Anyone who wants to know whether a pilot that only works in its own corner delivers anything should look precisely at that handover: not at what happens within one department, but at what happens the moment the work leaves the department.
Coordination between departments is a dependent layer. It only works once the fundamental matters are in place: an organization that knows who is responsible for what, an IT infrastructure that can connect departments, and data management that ensures that what one department records has meaning for the other. Without that foundation, every attempt at coordination between departments is a repair after the fact, and that repair costs more time than the application itself.
Whether your organization is ready for coordination between departments is not something you can read off the application itself. It depends on how the organization is structured, which systems already talk to each other, and whether there is someone who may decide across the boundaries of a department. The maturity measurement from hybridresourcing.com maps this out: five levels across seven dimensions, with a plot round in which multiple people from the organization score separately, so that the spread between departments becomes visible before that spread becomes a problem in practice. The tool is under construction; anyone who wants to make use of this can sign up for the waiting list.
This page describes what coordination between departments requires in order to be sustained. Which part of the work itself can be taken over by AI, per task and per department, is a different question. That is what the work scan from FTE TO AI calculates.
Vraag maar wat er moet staan voordat AI in uw organisatie kan landen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.