Judgement is now the main skill of journalists

Journalists' main skill was writing. Not any more.

AI can write, and increasingly it writes well. Agents can learn your style, transcribe interviews, interrogate hundreds of pages, find inconsistencies and test hypotheses.

So what's left? Judgement. Perhaps it always was.

What's the angle? What doesn't add up? Is the source credible? What's missing? Which fact matters? What question comes next?

Writing is the visible product

Behind every good story is a decision about what's worth writing.

AI can read the 200-page report. The journalist decides which paragraph matters.

It can suggest six questions for the chief executive. The journalist hears the hesitation after question six and asks the seventh.

It can find an anomaly in thousands of financial records. The journalist decides whether it's noise or a story.

Caution is justified. AI invents, misunderstands and produces plausible nonsense. Journalists must verify, and publishers must protect their credibility.

But there's a difference between protecting what you publish and restricting what journalists can learn.

The gap is widening

I talk to journalists almost every day, and I ask them about AI. Some are trailblazers. Many trail.

Those who experiment daily learn to frame problems, supply context, interrogate answers and spot hallucinations. Those who hold back fall behind, and so do their newsrooms. And their audiences.

Even trailblazers can't coast. The journalist who mastered AI two months ago and stopped experimenting is already slipping.

Leaders in other sectors are asking, "When should our agents be required to stop and ask a human?" Publishers should be asking: are our people learning to manage AI agents faster than our competitors?

Judgement becomes the premium

Routine production gets cheaper. Judgement gets more valuable. A first draft matters less. Knowing whether it's true matters more.

Reuters Institute has reported that technology companies are hiring journalists to train AI systems. What are they teaching the machines? Accuracy, clarity and structure.

Working with AI is collaborating with a robot. Weird, perhaps. But some of my best journalism was collaborative: producer and reporter, reporter and editor. AI is another collaborator.

That suggests two rules for newsrooms. Be conservative about what you publish with AI. Be adventurous about how your journalists use it.

Protect standards at publication. Before that, let people experiment, build agents, redesign workflows and fail safely.

Otherwise journalists lose as their skills fall behind. Publishers lose as newsrooms weaken. Readers lose when reporters do work machines can do instead of work only good journalists can.

So stop asking, "How can AI help me do this?" Ask, "Given what AI can do today, how should I do this now?"

Judgement without adaptation isn't enough.

Peter Wilkinson is the Managing Director of Wilkinson Group, a crisis communications and reputation management consultancy based in Sydney. His retained advisory service, Wilkinson Confidante, provides ongoing crisis preparedness and reputation counsel to boards and senior executives.