Automation arrives in newsrooms

My colleague John is a software engineer and was initially skeptical of AI. He now uses end-to-end AI automation to build software. His AI workflow retrieves issues from our ticketing system, creates an execution plan, writes the code, adds or updates automated tests, checks the code for correctness and formatting, and presents the code for his review in a pull request. John reviews the work at each step. Once he approves the pull request, that kicks off more automation, testing, then deploys the new software to production, and finally updates and closes the ticket.

That is the software development equivalent of starting with a story idea and having AI take it through concepting, research, synthesis, outline and first draft, revisions, copyediting, and then queuing it for publication.

The classic software tradeoff is that you either get speed or quality. With automation, my colleague John is moving 3 to 4 times faster and with similar or higher quality. He’s also having more fun. There are downsides. AI automates the drudgery that would have given him a mental break. The work that’s left is all heavy thinking. He has to meticulously review everything the AI creates, and he often goes home exhausted. Like the rest of us, he worries that AI might make us dumber in the long run.

Testing automation

Until now, most news staff have only used AI in isolated tasks, like searching through documents, summarizing, generating ideas, or proofreading. That’s going to change in 2026, as newsroom reporters and developers collaborate on end-to-end automation with human review. It will be easiest to experiment using flexible tools and custom code. Nothing is baked enough to add to your CMS yet.

Today, a news apps team could automate all aspects of publishing a data visualization: build a scraper, clean and categorize data, store it in a database, analyze and chart the data, write the application code, build the user interface based on a design mockup, and — after human review — deploy it. With some developer help, a city politics reporter might use AI automation to download government documents, transcribe town hall meetings, run AI queries against these primary sources, brainstorm several story angles, and even generate story outlines or first drafts. We’ll see bespoke automations for different beats and news companies of all shapes and sizes.

How far is too far?

Many news staff have been strongly against using AI to generate any part of a news article or script. There are valid concerns around copyright, accuracy, labor, and even mental wellbeing. Nonetheless, we’ll see more newsrooms experimenting with AI story generation in 2026. The easiest places to justify using AI in story development will be where newsrooms can serve audiences that would otherwise go unserved, or tell stories that would otherwise go untold.

We’ve had language translation and transcription powered by machine learning for a decade, and we’ve used it to translate existing stories and expand our audiences. AI advances make new applications of translation and transcription trivially easy by comparison, and these experiments won’t be controversial.

My team hit a tougher dilemma when building a blogging platform for small business owners. In the beginning, our team wanted no part of AI content generation. After several iterations and many heated debates, we now use AI to generate the first draft of blog posts. What changed our mind? Small business owners have great personal stories that never make it out of their heads, because they don’t have the time or expertise to write. They cede online storytelling and lose business to big company competitors. Still, the key to getting our product team comfortable with offering AI generation was seeing that small businesses could produce accurate, authentic drafts.

Journalists have weighed similar situations before. Our stakes with small business publishing may be different from what a typical news company faces, but our concerns and deliberations are similar to what you’ll encounter.

Go forth, but hold yourself accountable

Whether you pursue automations in engineering or storytelling, you will be uncomfortable and face difficult decisions. Write your principles down before you begin so that you can keep yourself honest. We routinely ask ourselves if we’ve made it easy for a novice writer to tell a personal story, but hard to generate spam.

Productivity gains are exciting. But quality and consistency are the foundations of trust. Build human review into all key decision points as you experiment. Humans are currently the best judge of quality, and that isn’t changing anytime soon.