AI in the newsroom:
co-pilot, not pilot.

I use AI across parts of my reporting and editing workflow, but I do not see it as an autonomous journalist. My rule is simple: co-pilot, not pilot. It can help with the heavy lifting. The reporter still decides what matters, does the reporting, checks every claim and owns the published work.

Start with the work that steals time

The most useful newsroom gains are often unglamorous. Tagging companies in a CMS, adding standard notices, transcribing an interview, sorting a long source document or preparing the same material for another platform can each take only a few minutes. Add those minutes across a breaking-news day and they become meaningful.

I saw this clearly while helping to train new reporters and editors. I brought together training logs, style guidance, recurring mistakes and the working knowledge that usually sits in an editor’s head. AI helped me turn that material into drills, onboarding prompts and more tailored feedback. The result was a living guide that could support someone joining a new desk. Mentoring remained human; the guide gave it a stronger starting point.

Use the saved time for reporting

Efficiency only matters if the newsroom spends the time well. I want fewer hours lost to repetitive production and more time spent picking up the phone, meeting sources and asking the follow-up question. AI cannot build a relationship with a source or notice the hesitation that changes the direction of an interview.

That human work is where stories gain texture and credibility. It is also where a journalist moves beyond the information everyone else has and finds something worth reporting.

A signal is a lead, not a conclusion

Financial journalism already has familiar reference points. Reporters know what to look for in an earnings release, a central-bank decision or an economic report. Newer markets often produce large streams of data without the same settled conventions.

AI can help organize those streams, surface patterns and flag movements that may deserve attention. That could give a newsroom an earlier view of a developing story instead of leaving it entirely reactive. The tool should stop at the signal. A reporter still has to establish why it matters, test other explanations, find knowledgeable sources and decide whether there is a story at all.

It can help me see what I missed

When I finish a piece, I want to know which questions remain unanswered. AI can map the themes already present in public coverage, compare them with my reporting and point to an angle that has received less attention. Sometimes that produces nothing useful. Sometimes it exposes the next question.

The value comes from the journalist knowing enough about the subject to judge the suggestion. Without that foundation, a plausible angle can easily become a weak or misleading one.

One piece of reporting can travel further

If I have spent days understanding a subject and interviewing an expert, the finished article should not be the only way readers can enter the work. The same verified reporting can support a short video or interview clip, a carousel or visual explainer, a social post or a follow-up question-and-answer format.

AI can help prepare those first cuts. Every version stays anchored to the same source material and receives a human edit. This is not about flooding every channel. It is about meeting audiences where they already get information without doing the underlying reporting five times.

Put the guardrails inside the workflow

A warning at the end of the process is not enough. I use a staged approach when working with a dense report or transcript. The instructions define the source material, the structure I need, the publication’s style and the boundaries the tool must not cross. Quotes and numbers are checked against the original. Uncertainty is flagged rather than filled with a confident guess.

• Use approved tools for unpublished, confidential or sensitive material.
• Keep the task tied to identifiable source material.
• Check every fact, number and quotation against the record.
• Apply house style, legal review and editorial standards as usual.
• Keep a named journalist or editor accountable for the final work.
• Explain material uses of AI when they affect what the audience receives.

A first draft can sound authoritative and still be wrong. The final human pass is not cosmetic. It is part of the reporting.

Newsrooms need reporters in the design process

AI adoption is also a culture question. A product team can build an impressive tool that solves no real newsroom problem. Reporters and editors need to shape the workflow, test it under deadline pressure and say clearly when it fails.

I prefer small experiments around a specific bottleneck. Test the result, listen to the people doing the work, improve it and invest further only when the value is clear. That approach also helps with the fear surrounding AI. People are more open to a tool once they can see it amplifying work they already do well.

The measure is better journalism

More output is not enough. I would judge a newsroom tool by tougher questions. Did it help the reporter spot a real lead? Did it create more time for sourcing? Did it improve the usefulness or reach of the work? Could the newsroom explain and defend the process?

This direction is consistent with the standards emerging across the industry. Reuters keeps editorial accountability with its journalists, while the Associated Press permits defined support tasks with human review. Reuters Institute research also finds audiences far more comfortable when a human leads and AI assists than when news is fully automated.

That is the newsroom I want to help build: curious about the tools, rigorous about their limits and determined to use any time saved for reporting that could not happen without a journalist. It is also the principle behind the editorial strategy and newsroom work I offer.