AI agents are no longer just for software developers. Tools like Anthropic’s Claude Code and OpenAI’s Codex, which started as coding assistants, are now being positioned to handle everything from booking travel to managing medical records. Kate Clark reports for The Wall Street Journal that the companies behind these tools see a multi-trillion-dollar opportunity in automating knowledge work for anyone who uses a computer.
The shift defines what many in the industry are calling phase two of the AI boom. The first phase centered on chatbots that answered questions. The second phase is about autonomous agents that take action, working for hours with minimal human oversight and learning user preferences over time.
“The long-term vision has always been a super-assistant,” said Nick Turley, head of OpenAI’s ChatGPT, “that can actually help you get things done.”
Coding tools set the stage. Cursor, a startup now valued at $29.3 billion, helped launch what insiders call the “vibe-coding” era, making it possible for non-engineers to build software using plain language. Claude Code and Codex followed, quickly generating billions in revenue. Claude Code alone accounts for $2.5 billion in annualized revenue, Anthropic has said. Codex reports more than two million weekly active users.
Boris Cherny, who leads Claude Code at Anthropic, frames the trend in blunt terms. “Coding is kind of the new literacy,” he said, “but luckily, it’s much easier to learn to code now than it was to learn to read, because you don’t have to practice.” He added: “I don’t want to sugarcoat it. It is going to be very disruptive.”
The disruption is already visible. Tens of thousands of job cuts have been linked to AI adoption, and a $1 trillion market selloff has followed as investors assess the technology’s potential to reshape finance, law and healthcare.
For OpenAI and Anthropic, the next frontier is winning over non-technical users, especially as both companies prepare for potential initial public offerings. OpenAI’s chief revenue officer Denise Dresser describes the moment in stark terms: “If you can think it, or you can describe what you want, you can build it.”
Power users are already living this reality. Venture capitalist Tomasz Tunguz runs dozens of AI agents simultaneously, routing his email, grocery shopping, travel research and news consumption through them. He estimates agents could generate $36 billion in annualized consumer revenue in the near future. Enterprise contracts, he says, represent an even larger opportunity.
Competition is fierce and financially unsustainable for now. Both OpenAI and Anthropic are charging far less than it costs to deliver their services, subsidizing usage to attract customers, much as Uber once offered discounted rides to capture market share.
Concerns remain significant. Users worry about data privacy, job displacement and security risks, including instances of agents deleting files or chatbots being linked to harmful behavior. Winning broader adoption will require the industry to address these issues directly.
