Anthropic Bets Big on AI Workers as Claude Takes On Full-Day Tasks

Anthropic is making an aggressive push into the next phase of artificial intelligence, unveiling new Claude models designed to do more than chat. The company says its latest systems can act more like digital workers, handling complicated assignments, using tools, searching the web, and carrying out tasks over long stretches with less human supervision.

The release marks a clear shift in strategy for the Amazon-backed AI startup. Instead of focusing primarily on chatbot conversations, Anthropic is now aiming squarely at AI agents — systems built to complete real work, not just respond to prompts. According to the company, its new Claude Opus 4 and Claude Sonnet 4 models are built to analyze large amounts of information, write code, generate polished content, and take action across different tools and data sources.

Anthropic says Claude Opus 4 can stay on task for nearly a full workday, a bold claim in an industry racing to prove that AI can move from assistant to operator. Both models are also designed to switch between reasoning and tool use, allowing them to search the web, work through long assignments, and preserve useful information from local files to maintain continuity over time.

Inside the company, executives are framing the update as a major leap forward. Chief Science Officer Jared Kaplan said Anthropic has been investing in these models for months, with a strong focus on improving Claude’s ability to handle demanding jobs such as research and coding. He also acknowledged the central challenge facing advanced AI systems: the more complex the assignment, the greater the risk the model can drift off course.

That tension is becoming one of the defining questions in the AI race. Every major player wants to build systems that can take over larger chunks of human work, but reliability remains a sticking point. Anthropic’s answer is to keep pushing Claude toward higher-value use cases while trying to prevent breakdowns in judgment or execution.

The company’s own product leaders say the new models are already changing how they work. Anthropic Chief Product Officer Mike Krieger said the latest Claude systems have crossed a threshold in writing quality, to the point that much of his own writing is now being generated by the model rather than simply shaped with its help.

The timing is no accident. Anthropic is growing fast, and investor appetite for AI remains fierce. The company confirmed last week that its annualized revenue hit $2 billion in the first quarter, more than double the prior pace. It also recently secured a $2.5 billion revolving credit line, giving it more financial firepower in an increasingly expensive AI arms race.

In plain terms, Anthropic is no longer just selling a smarter chatbot. It is selling the idea that AI can become a dependable coworker — one that can research, write, code, and execute tasks with limited oversight. Whether businesses fully buy into that vision may depend on one thing above all: whether Claude can keep producing results without going off the rails.