Put AI at the beginning, not at the end

Many newsrooms and publishers are looking for ways to use generative AI, and they’re making one crucial mistake: They’re hoping for the one magic button. They’re hoping for AI that can quickly turn an article into a podcast, web story, or newsletter. For AI that can quickly add illustrations to a product. For AI that allows untrained users to deliver results at the push of a button, within seconds.

That’s just not how it works (at least for now). Anyone who works directly and hands-on with AI will be able to confirm: In its current version, AI can help professionals with parts of processes. It can deliver first drafts, sometimes even spectacularly good ones, but it needs experts to work with the outputs.

An audio professional can create a good podcast faster with AI’s help. Because AI assists with rewriting and production. Someone who has written a text can create social media posts with the help of AI. The reverse is more difficult — if someone who doesn’t know the article uses AI to create social media posts, they need to compare the original and the output — and at that point, they might as well do everything themselves. A designer can develop an illustration style using AI and then generate hundreds of images and use a dozen of them. But it requires the practiced and trained eye; it requires art direction. This is not for novices.

For this to work, AI must be considered from the start, not at the end of a process. Otherwise, we’ll continue building our conventional processes and miss the opportunities that AI undoubtedly offers. This means examining all workflows carefully.

It also means staying realistic. Otherwise, disappointment is high when the automatically translated podcast needs a lot of post-production because the translations aren’t quite right, because timelines slip, or because the computer voice glitches. Or when a designer sees poor computer-generated images and despairs. A professional has the knowledge and tools to still achieve good results quickly. Everyone else will get frustrated.

In 2025, we need to consider AI from the start. Carefully examine all steps. This is where the opportunities lie. No one magic button, but many small magical buttons.