The only constant is change. For the past several years, much of the conversation around artificial intelligence has centered on generative systems—tools that can produce text, images, audio and video. These have captured the public imagination and, understandably, generated both excitement and concern across the industry. But a potentially more consequential shift is now beginning to take shape: the rise of agentic AI.
Before going further, let’s be clear about what this article is and is not. This is a thought experiment, not a forecast. The future described here may be a decade away, may look entirely different in practice or may not arrive at all. The value of the exercise is not prediction, but rather the perspective it can give about what is possible. As the saying goes: All models are wrong, but some are helpful.
With that framing in place, here is the central argument: the journalist of the future won’t directly write the story — they’ll train the agent that does. If that idea sounds like science fiction, read on. The pieces are already in motion.
What Makes Agents Different
Generative AI systems respond to stimuli (prompts). Agentic systems “act” and seek to accomplish broader objectives. Unlike a generative model, which responds to prompts, an agentic system can pursue goals, make multistep decisions and interact with other systems or agents on behalf of a person, organization or system. It is not so much a tool as it is a delegate.
Understanding this matters because it tells us where the real disruption lies. Generative AI changed what machines could produce. The evolution of agentic AI will change what machines can be trusted to do. The newsroom could be one of the clearest places we will see agentic technologies playing out in our industry.
If agents are assembling and delivering information to users without directing them to the original source, the traditional advertising and subscription models face obvious strain.”
The argument that follows rests on a specific assumption: that news will increasingly be assembled dynamically for each user, rather than consumed as static content. To be clear—not fabricated—assembled intelligently from trusted sources in response to what a user needs, when they need it and in the context of that moment. Whether that assumption proves correct or even desired is genuinely uncertain, but if true, it will change many roles.
There is already directional evidence of the possibility. AI-powered tools are generating summarized, citation-linked news experiences today. Users of search interfaces receive synthesized answers that draw from multiple publishers without directing them to any one source.
This is not entirely new—search engines have surfaced news snippets for years. What is changing is the sophistication of the synthesis and the degree to which audiences are satisfied without clicking through. The path toward dynamically assembled news is already being walked.
How News Is Encountered Now
If news is increasingly assembled by agents rather than read as discrete articles, then the journalist’s job cannot remain centered on writing those articles. And the current trends in how audiences consume news means that this future is more plausible.
Not everyone goes looking for news. Younger audiences, in particular, tend to encounter it while doing something else—scrolling through short-form video, moving through algorithmic feeds. News finds them; they do not seek it out. Push notifications and platform algorithms have become the primary editorial voice for a significant and growing share of the audience.
This matters because it means editorial control is already shifting—not to AI agents, yet—but to technology platforms and their recommendation systems. An agentic future would be an extension of a trend already underway. The cultural conditions for this shift are already forming.
One further implication would be that as news becomes more individually assembled, shared cultural experience erodes further. Monoculture—the common reference points that once came from everyone watching the same broadcast—have already shrunk dramatically. Agentic personalization would accelerate that fragmentation. The thesis we discuss carries consequences that are arguably negative.
What Changes in the Newsroom
If the journalist of the future trains the agent rather than writes the story, what does that look like?
The fundamental activities of journalism—interviewing sources, attending events, obtaining documents, verifying facts—are not going away. These are what give journalism its credibility, and they cannot be automated. What changes is that rather than assembling every story manually, a reporter feeds their gathered knowledge into a system trained on their expertise, voice and editorial standards.
Over time, that system—a digital extension of the reporter—responds to queries, synthesizes developments and surfaces context using what it has learned. The human journalist remains the source of credibility. The agent becomes the mechanism that scales it.
The implications ripple outward. The editor’s role shifts from revising individual pieces to governing the parameters within which agent systems operate. The news organization becomes less a publisher of discrete articles and more an operator of a trusted information system—one whose quality is determined not by today’s headline but by the integrity of everything that trained it.
This is not a reduction in the need for human expertise. It is a redirection of how expertise is applied. And it places an enormous premium on the thing that has always mattered most in journalism: the quality and credibility of the reporter behind the byline.
If agents are assembling and delivering information to users without directing them to the original source, the traditional advertising and subscription models face obvious strain. How publishers get paid in this world does not yet have an obvious answer.
Emerging ideas around usage-based or token-based compensation for content access are being discussed, but none has yet gained traction. These are not technical problems. The economic and governance problem may take longer to solve than the underlying technology takes to mature.
Why This Model Matters Now
The reason to think through this scenario today is not to prepare for an imminent transformation. It is to avoid being surprised by a gradual one. Technology shifts in media never arrive all at once—the transition from tape to file-based workflows took the better part of two decades—but the organizations that engaged early tended to fare better than those that waited for certainty.
The journalist of the future may not directly write the story; they’ll train the agent that does. If that future arrives, what will matter most is not whether AI is telling the news. It is who shaped the agent doing the telling, what they fed it and what standards they held it to. These are human decisions. They always will be. And the time to start making them thoughtfully is now.
Micky Silverman is a senior manager in Deloitte’s M&E practice who has worked for the past decade across traditional linear television providers, over-the-top media services, and major social media platforms, advising companies on their content strategy, product development and business operations. He can be reached via TV Tech.
