AI News/AI News

Anthropic's Duet agent offers useful work help, with some misjudged nudges

The Verge tested Claude Duet, Anthropic's consumer work agent, and found it good at spotting gaps across tools but weak on nuance and tone.

Key points

  • Duet is a research preview from Anthropic, open only to US subscribers on the Pro, Max or Ultra plans, and new users join a waitlist.
  • It targets professionals and small business owners, and decides on its own which items deserve attention first.
  • Once linked to email, calendar, Google Docs and Slack, it posts morning highlights and swipeable task suggestions, including drafted messages.

Why it matters for marketers: Agents that read across email, chat and docs can catch missed deadlines and conflicts, but tone misjudgments and auto-drafted writing mean marketers still need to review what goes out.

Anthropic has released Claude Duet, an AI agent aimed at professionals and small business owners who do not write code. It is labeled a research preview, with its own mobile app, and for now it is available only to paying US subscribers on three Claude tiers, with a waitlist for everyone else.

The Verge's reviewer connected work email, calendar, Google Docs and Slack for a week. Duet linked a point from a meeting summary email to a Slack message, inferred who the reviewer's boss was from a shared document, and opened each day with highlights. It also offered task cards that can be swiped to accept or dismiss. In one case it noticed that upcoming time off left little room before a draft deadline, then proposed a fix and drafted a Slack note to an editor, labeled as sent with Claude.

The weak spot was judgment. Duet pushed the reviewer about an edit still pending at an early hour in the editor's time zone, missing that a promised publish date was not firm. The reviewer also disliked its repeated offers to start drafts, preferring it as a backstop for organization rather than part of the creative process.

For marketers and content teams, the takeaway is that cross-tool agents are most valuable at catching overlooked conflicts between channels, and least reliable on tone and urgency. Anyone testing agents with client or editorial work should check drafts before they go out and consider how much inbox and chat access they are comfortable granting.

Anthropic says it is still testing how well Duet judges when to message users, so its nudges may improve. Teams on the eligible plans can join the waitlist and evaluate it on low-stakes workflows first.

Sources

This brief is our own summary. Read the full reporting at the source.

You read about AI automation. We run it.

DocFluence, built by Z Media, writes, designs, films and publishes for your brand every day, in your voice, from one login.