Anthropic unveils Claude Duet, an AI assistant that works on its own [Global AI Briefing]

[Financial News] Anthropic has unveiled Claude Duet, an artificial intelligence (AI) assistant that monitors users’ work and proactively suggests tasks they need to do, without requiring them to issue instructions one by one.
With Anthropic joining the AI assistant market after Meta Platforms’ Muse and OpenAI’s Dots, the global AI market appears to be expanding beyond chatbots that answer questions into a competition to develop AI agents that connect email, calendars, and work apps to handle real tasks.
On the 22nd (local time), Anthropic announced on its official blog that it would launch a research preview of Claude Duet, an AI agent service based on its next-tier AI model, Claude Opus 5.5.
The service is aimed at professionals and small business owners. It is available to Claude Pro and Max subscribers in the United States who gain access through a waitlist, and runs in a standalone app and on the web.
Claude Duet’s biggest feature is its proactive handling of tasks.

For example, every morning it selects and summarizes important information from connected apps, and checks for updates every few hours throughout the day to flag matters that need attention. It can handle everything from simple tasks such as scheduling meetings to preparing financial and sales forecasts, building dashboards, and developing web applications. It can continue working with its own virtual computer and browser even while the user is away.
It can also connect directly to workplace tools. Through at least 6,000 connectors in the Claude Marketplace, it can check messages in Gmail and Slack and draft replies. It can also look up outstanding payments in QuickBooks and Stripe and draft payment reminders, as well as assist with human resources and payroll management, writing job postings, and preparing electronic contracts. Business purchases can also be made using Stripe’s digital wallet Link and Shopify’s Shop Pay.
Although Claude Duet can proactively perform tasks without user instructions, it requires user approval for important actions such as sending messages, making payments, signing contracts, and deleting files.
Anthropic explained that Claude Duet cannot handle tasks such as stock trading or transferring funds between accounts, even with the user’s approval. It was designed so that it does not directly view passwords, and additional safeguards have been put in place to prevent sensitive information from being exposed. Anthropic added that the assistant may misjudge urgency or misunderstand conversations, and that it still has limitations when it comes to personal booking tasks.
AI assistants released by global AI companies each have different areas of focus.
Meta Platforms’ Muse targets everyday tasks such as shopping and managing calendars and email. OpenAI’s Dots helps users with their work across multiple workplace apps. While Meta Platforms is targeting its large user base, including Facebook users, and the market for services aimed at general consumers, OpenAI is effectively targeting workers’ working hours. Anthropic, by contrast, appears to be pursuing a strategy focused on improving the work productivity of professionals and small business owners.
With Anthropic entering the AI assistant market, competition in the global AI market is expected to shift rapidly from chatbots that answer questions well to assistant services that get work done on people’s behalf.
[email protected] Lee Gu-sun Reporter