Anthropic says Claude can now send emails through Gmail and manage files in Google Drive, turning its assistant from something that helps write work into something that can begin carrying it out. The feature is available on Claude’s paid plans, with users connecting Gmail or Drive through the connectors menu and retaining control over when approval is required.

For anyone who spends their day bouncing between email threads, shared documents, folders and half-finished replies, the appeal is immediately obvious. A request such as “reply to this client, keep it warm but firm, attach the latest proposal and save the final version in the project folder” has traditionally involved several manual steps. Claude’s newly announced capability aims to compress that sequence into a conversational task.

That is the real shift here. AI assistants have become competent drafting partners, but drafting has always been the easy part. The friction lives in the handoff: copying text into Gmail, selecting recipients, checking the right thread, finding the correct file, uploading it, deciding where a new document belongs, and then finally hitting send. By gaining the ability to act within Gmail and Google Drive, Claude is being positioned less as a writing tool and more as an operational assistant.

The announcement also lands at a time when the industry is racing toward so-called agentic AI – software that can not only answer questions, but use connected tools to complete bounded tasks. Anthropic describes connectors as integrations that let Claude access data and take actions inside third-party services, while inheriting the permissions of the individual who connected the account. That sounds technical, but the practical implication is simple: the useful version of an AI assistant is increasingly one that can work where the work already happens.

There is an important caveat, though. Anthropic’s own Google Workspace support documentation, updated in May, still says the Gmail connector can search and read email, organize threads and labels, and create drafts, but that sending must be done manually through Gmail. The newer Claude post explicitly says the assistant can draft and send replies, subject to user-controlled approval. That discrepancy suggests the capability may be a newer rollout, a plan-specific change, or a feature that has arrived ahead of its public documentation being fully updated. Users should therefore pay attention to the precise permissions and approval settings presented when they connect an account.

That distinction matters because email is not just another app. It is where contracts, invoices, customer relationships, hiring conversations, personal data and the occasional badly timed “reply all” live. Giving an AI permission to search a mailbox is one thing; letting it send a message in the user’s name is another. The difference between a helpful assistant and an expensive mistake can be a single overly confident sentence sent to the wrong person.

Anthropic appears aware of that concern. Its announcement emphasizes that the user controls when Claude needs approval, rather than presenting sending as an entirely autonomous background process. In the broader connector framework, administrators can set action permissions to “always allow,” “needs approval,” or “blocked,” separating read-only access from more consequential write or deletion actions. For businesses, that approval layer may be the feature’s most important part, not an annoying extra click.

Drive management is arguably the quieter but more broadly useful half of the update. Modern work is often buried in file sprawl: “final” documents that are not final, folders that only one person understands, PDFs attached to emails instead of saved centrally, and a pile of meeting notes nobody will revisit. Claude’s Google Workspace integration already supports looking up Drive file metadata, reading common file formats, uploading files, creating folders, reviewing file permissions and saving Claude-generated files to Drive when file creation is enabled. In that context, a more action-oriented workflow could make Claude valuable for the tedious digital housekeeping that tends to get postponed indefinitely.

For a small business owner, the potential use cases are easy to imagine. Claude could prepare a reply to a supplier based on a prior email thread, create a follow-up document from the discussion, place it in the relevant Drive folder and keep the human in the loop before any message goes out. For a content team, it could find the latest brief, pull context from supporting documents, draft a response to an editor and save a revised version in the shared workspace. The gain is not necessarily that the AI writes better prose than a person. It is that the person spends less time shuttling information between systems.

Still, the feature should be treated as an assistant, not an autopilot. Email has more hidden context than it first appears: office politics, customer history, legal commitments, cultural tone and the tiny details that distinguish a friendly follow-up from a message that accidentally escalates a problem. AI can summarize a thread, but it may miss what was said over a phone call, what a recipient dislikes, or why a deliberately vague sentence was left vague.

The safest early use will likely be structured, low-risk work: drafting replies to internal threads, sending routine status updates after a review, filing generated materials into clearly defined Drive folders, or responding to common customer requests with approved language. High-stakes tasks – contracts, payments, personnel matters, sensitive client disputes and anything involving confidential information – deserve a human review every time.

There is also a broader competitive story behind the feature. Anthropic has been steadily building Claude’s connector ecosystem beyond simple chat. Its support materials describe connectors that can reach services such as Google Drive, Slack and Linear, while the Model Context Protocol gives developers a common way to connect Claude with outside tools and data sources. Meanwhile, Anthropic recently expanded Microsoft 365 support so that enabled organizations can let Claude draft, send and organize Outlook email, manage calendar events, and create or update OneDrive and SharePoint files.

In other words, Gmail and Google Drive are not isolated additions. They are part of a contest over which AI assistant becomes the layer sitting above a company’s everyday software. The chatbot window matters less than the ability to understand a user’s work, retrieve the right context and safely take the next action.

Claude’s Gmail and Drive move is promising precisely because it focuses on the messy middle of knowledge work, where most time is lost. But it will earn trust only if its permissions are legible, its approval prompts are reliable, and users can clearly see what it plans to do before it does it. Sending an email is easy. Sending the right email, with the right attachment, to the right person, at the right moment is still very much a human skill.