If every company's AI agents think alike, what is left to set them apart?

A Monday morning with an agent swarm

The essay, published by Noema, opens with Lukas Weber, a 28-year-old copywriter at a mid-sized software company. Each Monday he starts by checking on the AI agents he has built. An agent is an AI program that carries out a task, such as drafting a newsletter or answering email, with little step-by-step prompting.

One agent gathers industry news and customer calls to suggest newsletter topics, then passes them to a writing agent. Another sends tailored proposals to prospective clients who visit the website. A third flags worrying trends in campaign data. Weber says his bosses expect him not to write at all. He reviews and approves what the agents produce, and his team has shrunk this year.

Weber's colleagues in HR, sales, customer success and engineering work the same way. "It's how everyone works now," he says. The author, Aaron Horwath, treats this as the picture of success that AI-enthusiastic executives describe.

The problem: one brain, many companies

Horwath's worry is that competitors are doing the same thing, on the same small set of large language models (LLMs, the text-predicting systems behind chatbots like ChatGPT and Claude). In his telling, it is as if every firm in an industry had hired the same single employee.

He calls the result "airbrain," a play on writer Kyle Chayka's "Airspace." Chayka described how social media algorithms pushed coffee shops, bars and hotels in different cities toward one look: minimalist furniture, reclaimed wood, industrial lighting. Horwath argues that newsletters, websites and ads are now converging on a shared AI tone in the same way.

He points to research on this. As the essay describes it, Emily Wenger of Duke and Yoed Kenett of Technion tested 22 commercial LLMs against more than 100 people on standard creativity tasks. Individual models sometimes beat individual people. But as a group, the models landed on very similar answers, while the people did not.

What the Kissimmee River has to do with it

To explain why efficiency can backfire, the author tells the story of a Florida river. The Army Corps of Engineers saw 103 miles of bends in the Kissimmee as a flooding problem. Between 1962 and 1971 it straightened the river and dredged it to 30 feet deep.

The flooding stopped, but the river's natural ability to trap silt was lost. Nearly 35,000 acres of floodplain habitat were destroyed, and roughly 90% of the wintering waterfowl population was wiped out. Work to put back 44 miles of bends began in 1999 and finished in 2021, at a cost of more than $1 billion.

Horwath sees a parallel inside companies. The old way of working was slow: someone has an idea, colleagues critique it, it gets refined, then it gets funded. AI lets one person skip most of that and build alone in hours or days. But those meandering steps, he argues, are where an idea picks up what makes it distinctly yours. A company's edge is the combined talent of the people it hired, and routing everything through AI filters that out.

Skills may fade, and the edge with them

The essay cites Jay Barney, a strategy professor at the University of Utah, and his co-authors in MIT Sloan Management Review. They argue that AI can give only a short-term advantage, because everyone gets access to the same value. Their view is that it will make companies more alike, not set them apart.

There is also a human cost. Horwath cites a 2026 Nature report finding that skills can fade quickly once AI is introduced. In a 2026 survey of 2,500 employees, 39% said relying on AI was making them less intelligent. His warning: a company that later wants to return to human-made work may need time to rebuild the skills it let slip.

The essay's final section, as far as the text we have shows, turns back to coffee shops. When cafés all looked alike, researchers found people struggled to tell them apart. In one test, only 6% could tell Chicago from San Francisco. Then sameness stopped paying, and cafés began competing on distinct experiences. Horwath expects knowledge-work firms to reach the same conclusion. He says the answer is not to drop AI. Some firms should lean on it harder, and others should restore some of the winding human steps they cut.

How to read an essay like this

This is an opinion essay, as noemamag.com published it, not a controlled study. It leans on research and anecdote, and its central claim, that heavy AI use erodes competitive edge, has not been tested at scale in the material we have. Weber's story is told as one example.

Still, the question it raises is practical. Which of your tasks does AI make faster, and which does it quietly make more generic? The answer is likely to differ by person and by job, and you can check it yourself.