Agentic AI in publishing: automating processes, staying in control

Agentic AI systems can already do far more today than generate individual texts or summarize information. They research, process content, access different systems, and carry out multiple steps in sequence. In publishing too, this opens up new possibilities, for example in research, content preparation, or quality assurance.

But with growing automation, the way we handle control changes as well. Not every step should run fully autonomously, and not every decision can meaningfully be handed off to an AI agent.

That's exactly what Matthias Kraus will address on September 16 at ai@media, the Digital Publishing Report's conference on the practical use of AI at publishers.

Agentic AI in practice

The talk shows how agentic AI can be integrated into existing publishing processes while keeping transparency and control intact. What tasks can an AI agent handle on its own? Where does it need human approval? And how do individual steps stay traceable?

Using a concrete example with Langdock and Xpublisher, Matthias Kraus shows what that can look like in practice. Langdock provides the foundation for custom AI applications and agentic workflows, while Xpublisher, as the editorial and publishing system, takes over further processing and publication of the content.

Research and content preparation, for example, can be supported by AI, with the results handed directly to Xpublisher. There they're reviewed editorially, refined further, and published through existing publishing processes.

Not every step needs the same level of control. Clearly defined tasks can run automated, while decisions with editorial or legal weight stay with people.

Who should attend

The talk is aimed at editorial, product, production, and IT leaders who want to use AI for more than isolated tasks and integrate it more deeply into their existing processes.

The focus isn't on theoretical future scenarios, but on concrete approaches for using agentic AI in publishing without giving up transparency and control.