The editorial pipeline at a university communications operation moves through many hands. Communicators across departments — each with their own priorities, schedules and varying levels of communications experience — are responsible for generating content that ultimately represents the institution. Getting that content to a consistent standard, on deadline, is a structural challenge most university communications teams know well.
At San Diego State University, I saw an opportunity to move the feedback earlier in the process.
I designed and implemented an AI editor scoped specifically for our newsroom environment. The tool checked copy against our internal style guide, flagged structural issues, assessed story length and provided guidance on the basics of strategic communications — before a story ever reached an editor’s desk. The goal wasn’t to replace human judgment. It was to give communicators a first reader that could help them cover the fundamentals on their own timeline, before the editorial team needed to weigh in.
The framing mattered as much as the tool itself. We were deliberate about how we introduced it: this isn’t a crutch, and it isn’t here to make decisions for you. It’s support for your workflow. That distinction landed. The response was positive. Communicators used it not just to check their work but to learn — story structure, format, length, style — in real time, on their own terms. For many, it was the first time they’d received that kind of immediate, specific feedback outside of a formal training session.
The editorial team recovered time for higher-value work. Communicators grew more confident in the fundamentals. The tool did what the best editorial infrastructure does: it raised the floor without lowering the ceiling.
What the Industry Is Learning
The SDSU experience reflects a pattern emerging across professional newsrooms. Where AI adoption has worked, it hasn’t been because organizations found a way to automate journalism. It’s been because they found specific, bounded problems the technology could support without displacing the judgment that makes journalism worth reading.
The Associated Press has used automation for years to generate routine financial and earnings reports, freeing journalists to focus on higher-value enterprise and investigative work. ProPublica uses AI techniques to analyze large datasets — public records, satellite imagery — surfacing patterns and leads that would be difficult to find through manual analysis alone. The New York Times has employed AI to generate article summaries and support translation efforts. Norway’s public broadcaster NRK reported increases in time spent on page after introducing AI-generated bullet-point summaries.
None of these are examples of AI replacing editorial judgment. They’re examples of AI handling specific, well-defined tasks so human judgment can go where it’s actually needed.
The Boundary That Makes It Work
The distinction that matters most isn’t technical. It’s philosophical.
AI tools positioned as support systems — assistants to human judgment rather than substitutes for it — tend to get adopted. They lower friction, build confidence and compound over time. Tools positioned as autonomous creators, or deployed without clear governance, tend to generate resistance, workarounds and the kind of informal “shadow AI” use that introduces ethical and legal risk without any of the oversight.
Newsroom leaders who have navigated this well share a few things in common: they define what the tool does and doesn’t do before it launches, they invest in staff literacy alongside the technology, and they treat AI systems with the same editorial scrutiny they’d apply to any other part of the workflow.
What This Means in Practice
The long-term impact of AI on journalism won’t be determined by the technology itself. It will be determined by how news organizations choose to deploy it, govern it and talk about it with their teams.
The organizations getting this right aren’t the ones moving fastest. They’re the ones being most intentional — treating AI as infrastructure, not innovation for its own sake, and keeping human editorial authority exactly where it belongs.
The tool we built at SDSU wasn’t revolutionary. It was useful. And in a deadline-driven environment, useful is exactly what it needed to be.
