What jumped out to me is how aggressively practical this is. Anthropic is not pitching Claude Duet as a flashy consumer toy or a vague “future of work” thing. It’s clearly going after the boring, valuable stuff: briefs, triage, scheduling, dashboards, forecasts, web apps. That’s the kind of workload where people will forgive a lot of ugliness if the agent reliably saves them an hour.
I’m still a little skeptical, though, because the gap between “can do this in a demo” and “I’d trust it with my calendar, inbox, and connected tools every day” is enormous. Anthropic’s own caveat — that it can misread a thread or miscalibrate urgency — is doing a lot of work here. That reads less like a footnote and more like the core product risk. If an agent is meant to interrupt you only when something matters, getting urgency wrong is basically the whole game.
The part I’d actually want to test is the approval flow around sensitive actions. That’s where these systems either become genuinely useful or stay stuck in novelty land. If it really pauses before payments or deletions, and if the classifier does a decent job of refusing strict sensitive data like credit card numbers, then maybe this is closer to a usable assistant than the usual “let it loose and hope” agent story. But Anthropic is also admitting it’s an early version with uneven performance, especially on personal errands. That makes sense. Buying tickets, dealing with messy consumer websites, and handling real-world exceptions is where agents tend to get embarrassing fast.
I also think the target audience matters here. “Professionals and small business owners” is a very specific sweet spot: enough repeated admin pain to justify an assistant, but not so much compliance baggage that the product becomes impossible to ship. The standalone mobile app plus web access suggests Anthropic wants Duet to feel like a companion layer on top of Claude, not a separate universe. That seems smart.
Still, the naming and positioning make this look like part of a broader race to own the “agent” category before the market settles on what an agent even is. Meta has Muse, OpenAI has Dots, now Anthropic has Duet. The branding is almost embarrassingly human and friendly, which maybe says something about the strategy: make the thing feel like a collaborator, not a bot. Whether users buy that depends on one boring detail above all others — whether it stays out of the way until it’s actually useful.
Reference: Anthropic takes on Muse with Claude Duet, its new AI agent
