Anthropic Launches Claude Duet AI Agent Targeting Professionals and Small Businesses

Anthropic announced Claude Duet on 21 October 2026, introducing what the company describes as a proactive AI agent designed to handle background tasks for professionals and small business owners — a step beyond the conventional prompt-and-response model that most AI tools currently follow.

The product is available as a standalone application on iOS and Android, and through claude.ai. Access at launch is limited to US-based Claude Pro and Max subscribers, who can join a waitlist to try the research preview. No separate charge applies; the preview is included within existing Pro and Max subscription tiers.

What Claude Duet Actually Does

The core proposition is straightforward: rather than waiting to be asked a question, Duet monitors applications connected to it, identifies items that need attention, and either prepares or completes assigned tasks while the user is elsewhere — including while their computer is switched off, using a virtual machine and browser to continue working in the background.

Anthropic says users can expect a morning brief summarising priorities drawn from their connected tools. The agent can also contact users when something in a connected application requires a decision.

Reported use cases include scheduling meetings, building dashboards, preparing financial or sales forecasts, and prototyping or developing web applications. One secondary technology publication reported that Duet can connect to more than 6,000 applications, though that figure has not been independently verified against an official Anthropic source.

The approval requirement is the design element Anthropic has placed most emphasis on publicly. Before sending messages, making payments, signing documents or deleting information, Duet must obtain explicit user approval. That safeguard is built into the preview by design, not as an optional setting.

The Competitive Landscape

Anthropic is not alone in this space. Claude Duet enters a developing AI-agent market alongside products from OpenAI and Meta, both of which have moved in the direction of automated knowledge-work tools. The race to build agents that do things, rather than simply answer questions, has accelerated considerably across the industry in 2026.

But Anthropic’s positioning leans into caution. The company has been open about the fact that Duet remains experimental, that the agent has limitations, and that human oversight is still necessary. That framing sets a particular expectation for early adopters.

Limitations and Open Questions

The research preview label matters here. This is not a finished, generally available product; it’s an early-stage release to a limited pool of US subscribers, with broader rollout timelines unconfirmed.

Several questions remain unanswered. Team and Enterprise availability has not been confirmed. UK launch timing, UK pricing, and any applicable UK regulatory arrangements — including guidance under UK GDPR — have not been verified. For UK businesses weighing up AI agents that connect to sensitive business systems, those are not minor details.

There are also practical risk questions that any organisation would need to work through before deploying an agent of this type. An AI connected to business applications and capable of acting within them creates potential exposure around privacy, data security, accuracy of automated outputs, and accountability if something goes wrong. The approval-gate mechanism reduces some of that risk but doesn’t eliminate the need for careful human review of everything the agent produces or proposes.

Reactions and Industry Context

Dario Amodei, Anthropic’s chief executive, has previously described the company’s direction as building AI that can take on complex, multi-step work rather than single-turn tasks. Claude Duet is a concrete expression of that direction, though the research-preview framing signals that Anthropic is treating this as a learning exercise as much as a product launch.

The small-business angle is worth paying attention to. Scheduling, forecasting, dashboard preparation, and customer follow-ups are exactly the kinds of tasks that consume time in small organisations that don’t have dedicated administrative staff. Whether Duet delivers on that promise at scale — and at what reliability — will become clearer as the waitlist opens up and more users report their experiences.

And the approval-before-action model, while sensible, does raise a question: if a user still needs to review and approve every consequential action, how much time is actually saved compared with doing the task directly? That’s an empirical question the research preview period is presumably designed to answer.

What This Means for Kent Residents

At launch, Claude Duet is not available to UK users; access is currently restricted to US-based Claude Pro and Max subscribers via a waitlist, and UK availability, pricing and regulatory arrangements have not been confirmed by Anthropic. Kent professionals and small business owners who already use Claude for day-to-day work should watch for any UK rollout announcement, chiefly given the potential for tools like this to reduce routine administrative workload. Anyone considering future use of AI agents connected to business systems — whether in a small firm, a council department, or an NHS setting — would need to review data protection obligations under UK GDPR before granting an automated tool access to sensitive applications or communications.