Use Case: Embedded AI Lead in a Publishing Company

A publishing company usually does not have one neat “AI problem”.

It has a collection of smaller things happening at the same time. A SharePoint structure needs to be redesigned. Departments have different ideas about how information should be stored. Marketing data could be moved automatically instead of being prepared manually. People are experimenting with AI, but usage differs from team to team. Management has ideas that need to be investigated before they become projects.

This is the environment we work in.

We provide an embedded AI Lead who works with management and the individual departments, understands what is already happening and helps move the useful initiatives forward.

The role is intentionally broader than AI.

If somebody asks for an AI solution, the first question is usually what they are actually trying to achieve. Sometimes AI is appropriate. Sometimes a normal automation is enough. Sometimes the underlying process or information structure needs to be fixed first.

That distinction saves quite a lot of unnecessary building.

For example, part of our work with the Publishing Company has been around SharePoint information architecture.

At first glance, this sounds mainly technical. In practice, the difficult questions are organisational.

What information belongs where? Who owns it? Which departments need access? Which metadata is useful enough that people will actually maintain it? What should be migrated, and what has simply accumulated over time and does not deserve a new home?

We work through these questions with the departments in workshops and turn the answers into something that can actually be implemented.

The technical SharePoint structure is only one result. The more useful result is that ownership, access and the way information should be handled become clearer.

The same pattern appears in automation work.

We have worked on connecting marketing information with Power Automate so that reporting does not depend entirely on somebody collecting and moving the data manually.

The interesting part is not that two systems can technically exchange information. That is usually the easier part.

We also need to understand which information is required, when it should move, what happens when something is missing, who notices when the automation fails and what the resulting report is supposed to support.

Otherwise we have successfully automated a poorly understood process, which is efficient in a fairly narrow sense.

AI adoption works in much the same way.

Employees are already exposed to AI tools. Some experiment independently, some avoid them, and different departments naturally find different uses.

Our role is to help the company turn this into something more deliberate.

That includes working with departments to understand where AI could help their normal work, demonstrating practical uses, helping employees understand the limitations, and supporting management with decisions around approved tools, information handling and sensible internal rules.

The objective is not to increase the number of AI tools being used.

It is to find situations where using one produces a better result than the current way of working.

This is also why being embedded matters.

If we were commissioned separately for every small initiative, we would repeatedly have to learn the organisation from the beginning.

By staying involved, we accumulate context.

A discussion with one department may explain why another department is struggling with the same information. A SharePoint decision may affect a later automation. An AI idea may depend on data that we already know is inconsistent because of another project.

Those connections are difficult to see when every piece of work is treated as a separate supplier assignment.

The working method is therefore fairly practical.

We talk to management and departments, identify problems and ideas, clarify what is actually happening, and decide what deserves further work.

Some topics become projects.

Some become small prototypes or automations.

Some require a workshop or a clearer process.

And some stop after investigation because the expected benefit does not justify doing more.

That last category matters as well. Digital transformation becomes expensive quite quickly if every idea automatically becomes a project.

The embedded AI Lead also gives management one place where these initiatives can be kept visible.

What are we currently working on? What is blocked? What needs a decision? Which idea should be investigated next? Which initiative no longer makes sense?

That creates continuity between management decisions, departmental requirements and technical implementation.

The Publishing Company still owns the business decisions and the subject knowledge. We are not there to tell editors, marketers, salespeople or management how publishing works.

Our role is to understand enough of their work to see where technology can remove unnecessary effort, improve access to information or make something easier to manage, and then help turn that opportunity into something usable.

Sometimes the result is AI.

Sometimes it is automation, a better information structure, a small internal tool or simply a clearer way of working.

The point is not to force all of those problems into the same technology.

The point is that somebody is responsible for finding out which solution makes sense and then moving it forward.