AI Firms Race to Capture Office Work with Agents

Anthropic introduced Claude Duet on Wednesday, Oct. 21, an AI agent for work that can read users’ email, calendars and business software, then act across more than 6,000 outside apps. The tool is in an early test phase, available through a wait-list to U.S. subscribers on Anthropic’s Pro and Max plans, with no extra charge beyond their subscription fee.

A product built around work

Claude Duet sends a morning briefing drawn from a user’s email, calendar and business software. It checks those accounts every few hours, then contacts the user if it finds something that may need attention. Anthropic says the agent takes its Claude chatbot beyond answering questions, allowing it to handle parts of users’ working lives.

The tool connects to outside services such as Canva and DocuSign. It can also make purchases through Link, Stripe’s digital wallet, or Shopify’s Shop Pay, but users must approve each one. That brings workplace information and commercial transactions into a single assistant, while leaving purchase approval with the user.

The reported details focus on what the agent can access and do. Anthropic says Claude Duet will show users exactly what it plans to do before it sends a message or makes a payment. A new filter aims to stop it from sharing credit-card numbers or government ID numbers. The company has also blocked some actions, including stock trading and moving money between accounts, even when users approve them.

The competition for workplace agents

Anthropic said Claude Duet competes with OpenAI’s dots, which debuted the previous month, Meta’s Muse, Google’s Gemini Spark and agents from start-up Instinct. Instinct raised $1 billion in its latest funding round. Those figures and product names point to a crowded field, as companies race to build assistants that work across users’ digital accounts and business tools.

For U.S. users, a paid subscription is the way to join Claude Duet’s wait-list: customers need Anthropic’s Pro or Max plan, and the agent costs nothing beyond that subscription. The article doesn’t give a separate price for Claude Duet or report product revenue. It also doesn’t explain how any money from purchases through connected services gets divided.

Controls and what remains unreported

Anthropic’s stated safeguards cover specific actions and sensitive information. Users must approve purchases, the agent previews messages and payments, and it can’t carry out certain financial actions. Still, the account-checking feature connects the agent to email, calendars and business software. The article gives no further details about how long account information stays on file or how the company handles it.

The article describes a work-focused product and its safeguards, but offers no figures on jobs, wages or changes to workers’ workloads. It reports no union response or organized labor position, and no government action or broader policy proposal. Instead, the available information shows the commercial contest: Anthropic is offering a tool that can monitor connected work accounts, flag matters needing attention and carry out approved actions across more than 6,000 outside apps.