Why newsrooms keep reaching for the wrong AI tool

Would you drive a Formula 1 car to pop to the shops for a pint of milk?

It sounds like a silly question. Nobody would. It's absurdly expensive, wildly impractical, and completely overpowered for the job. And yet, walk around almost any newsroom experimenting with AI right now and you'll find exactly that happening, just with software instead of cars.

We raised this on a recent episode of The Signal, and it's stuck with us since, because once you start looking for it, you see it everywhere: reporters firing up the biggest, most expensive, most capable model available for tasks that a much smaller, cheaper, faster tool would handle just as well. Nobody's doing anything wrong, exactly. They're just reaching for the tool that's top of mind, not the tool that's right for the job.

The default is always the biggest hammer

There's a simple reason this happens: the flagship model is the one everyone's heard of. It's the one in the headlines, the one a colleague recommended, the one that's already open in a browser tab. Reaching for it isn't a strategic decision. It's the path of least resistance.

The trouble is, capability and cost move together. The models that can handle genuinely hard, ambiguous, high-stakes reasoning; untangling a dense document, spotting inconsistencies across a large dataset, drafting analysis that needs real judgment, are also the slowest and most expensive to run. Using that same tool to summarise a press release or clean up a transcript is like paying for a chauffeur to nip to the corner shop. It works. It's just a wildly inefficient way to get there.

Why this matters more than it looks like it should

On any single task, the cost difference between the right-sized tool and the oversized one is small enough to ignore. A few extra pence, a few extra seconds. Nobody notices, and nobody complains.

But newsroom AI use doesn't happen once. It happens dozens of times a day, across every desk, compounding quietly in the background of a subscription bill or an API invoice nobody scrutinises line by line. Multiply a small inefficiency by a whole newsroom, every day, for a year, and it stops being a rounding error and starts being a real, avoidable cost. One that's almost invisible until someone finally asks where the AI budget actually went.

That's the same blind spot we wrote about in how to measure the ROI of AI in your newsroom: most newsrooms are tracking whether a tool saves time, but almost nobody is tracking whether it's the right tool for the time it's saving. Both questions matter, and only measuring the first one hides exactly this kind of cost.

Bigger isn't just expensive. It's sometimes worse.

Larger, more powerful models are often tuned for depth, nuance and open-ended reasoning, which sounds like an advantage until you ask them to do something narrow and mechanical, like reformatting a list or extracting five facts from a document. Sometimes a smaller, more tightly scoped tool actually produces a cleaner, more predictable result, precisely because it isn't trying to be clever about a task that doesn't need cleverness.

In other words: the expensive option isn't a safe default. It's just a different set of trade-offs, and newsrooms are rarely choosing it on purpose.

A rough way to think about it

You don't need a procurement committee to fix this. You need a habit of asking one question before reaching for a tool: how much genuine judgment does this task actually require?

Tasks that involve real ambiguity; interpreting nuance, weighing conflicting information, producing analysis someone will publish under their name are worth the heavyweight model and the cost that comes with it. Tasks that are mechanical, repetitive, or narrowly scoped; transcribing, summarising something short, tidying formatting, very rarely need it. Most newsroom AI use, if you're honest about it, sits much closer to the second category than the first.

This is also where governance and cost efficiency turn out to be the same conversation. As we explored in our piece on shadow IT in the newsroom, a lot of tool choice happens invisibly, one reporter at a time, with nobody stepping back to look at the pattern. Getting a handle on which tool is being used for which task isn't just a budget exercise. It's the same visibility problem that sits underneath almost every AI governance question.

The fix isn't fewer tools. It's the right one, more often.

None of this is an argument for using AI less, or for banning the powerful models. They're genuinely worth it for the tasks that need them. It's an argument for treating tool choice as a decision, not a default. The newsrooms getting real value out of AI aren't the ones with access to the single best model. They're the ones who've built the habit of matching the tool to the task, every time, instead of reaching for whatever's already open.

It's a small habit. It's also, quietly, one of the highest-leverage changes a newsroom can make, both for the budget, and for the quality of what actually gets published.

This idea started life on The Signal, Episode 11, and connects to our pieces on measuring the ROI of AI in your newsroom and Shadow IT in the Newsroom. If you're not sure where your AI budget is actually going, get in touch - it's usually hiding in plain sight.