Claude Co-Work, a corporations AI tool unveiled by U.S. artificial intelligence (AI) developer Anthropic, has shaken the structure of the traditional software market and landed at the center of debate.
According to industry sources and foreign media on Feb. 5, Anthropic recently launched Claude Co-Work, a paid service that lets office workers without programming knowledge build workflow automation tools simply by talking with AI. It enables immediate app-style implementation of repetitive tasks such as document summarization, data analysis, and contract review without a separate development process.
Since the debut of Claude Co-Work, concerns have grown in the software industry that general-purpose AI could replace expensive professional software such as tax processors and even corporations services. Conventional software, once adopted, is hard to replace and has ensured stable subscription revenue, but analysts say AI-based automation tools could disrupt this structure.
The Financial Times said Claude Co-Work is overturning the very concept of traditional software development. Anthropic, a company that has focused on corporations AI rather than consumer chatbots, drew attention last year by releasing Claude Code for developers.
Claude Code spread the so-called "vibe coding" trend of writing code through conversation with AI, but the command-based environment and developer review process left a barrier to entry. In contrast, Claude Co-Work greatly improves convenience by letting users converse with AI directly in a web browser to build automation tools and instantly link them with workplace productivity software such as Slack.
Anthropic's strong coding capability is cited as the background for this competitiveness. Anthropic has focused on "AI feedback-based Reinforcement Learning," in which AI evaluates the results of another AI. Unlike the conventional method where humans provided feedback directly, AI repeatedly grades and improves on its own, boosting performance. This approach is known to be particularly effective in areas with relatively clear answers, such as coding or math.
Anthropic has said that AI currently generates 70–90% of its system code. It also open-sourced a technology called the Model Context Protocol, making it easy to connect apps built with Claude Co-Work to existing workplace systems. Various plugins based on this are growing rapidly.
A legal workflow automation plugin unveiled early this month drew particularly strong reactions, even affecting the stock prices of established legal software companies.
Caution remains, however, over whether Generative AI can replace large swaths of software. Because of the nature of Generative AI, the risk of "hallucinations," which produce plausible but incorrect results, persists, and experts note this can be fatal in error-sensitive fields such as finance and law.